skill

aos

A Curated AI AGENT OS – Optimized agent skills and add-ons like memB, OpenWiki, Heimdall Token Saver, and Godmode architectures.

hybridlabor-api/aoscommunitylistedunreviewed

Use skill

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/hybridlabor-api/aos itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 17, 2026

Language / Sprache / Idioma: English · Deutsch · Português

AOS — BDB Agent OS

AOS installs a curated skill library, a subagent roster, gate hooks and a runnable multi-agent build pipeline into every coding-agent harness on your machine.

npx -y @hybridlabor-api/aos@latest

Recommended: the full installer above. Alternatively, use the skills CLI:

Skills use the open Agent Skills format (SKILL.md). AOS installs them for Claude Code, Google Antigravity, Codex CLI, OpenCode, Cursor, Windsurf, Roo Code / Cline, Aider and the AOS CLI, so all of them behave the same way.

Why AOS

	Harness default	With AOS
Skills	Each harness has its own skill folder and its own copies.	270 skills in eight categories, written as `<name>/SKILL.md` into every harness you have. One source tree.
Subagents	Agent files written by hand, per harness, in that harness's format.	45 subagents compiled into each harness's native agent format (Claude Code, Antigravity, Codex, OpenCode only; others use rules in AGENTS.md).
Release safety	A rule in a prompt that the agent is asked to respect.	A `PreToolUse` hook (Claude Code, Antigravity, Codex, OpenCode) or rules in AGENTS.md (others) that blocks `git push`, `npm publish`, `npm version` and recursive `rm` until your own message is the literal word GO.
Multi-agent builds	Agents that call agents, with the routing inside their prompts.	A dispatcher graph: nodes never invoke each other, state is durable, a repair loop has a no-progress guard, escalation goes to you.
Plans	Scroll back through the chat to find what was agreed.	Plan Canvas: annotate the plan in a browser, then approve it before any agent builds.
Visibility	Read logs in several terminals.	agenttrail live map per repo, and a fleet band in Claude Code with gate mode, token-weather and who waits for your GO.
Sessions	Each session is alone in its terminal.	A2A between live Claude Code, OpenCode, agy and Codex sessions on localhost. An inbound message is never a GO.
Beyond code	Coding assistants know code.	19 media-eventtech skills and MCP servers for grandMA3, Resolume, TouchDesigner, Unreal, DaVinci, After Effects, Blender and Rhino; `godmode-hardware-pcb` for KiCad and OpenSCAD; playbooks for offers, invoices, crew call sheets and event trackers.

What you get

See Playbooks for guided workflows across these domains.

Quick start

I want to...	Run
Install for every harness I have	`npx -y @hybridlabor-api/aos@latest`
Install with low context overhead	`npx -y @hybridlabor-api/aos@latest --profile=minimal`
Refresh only the gates and hooks	`npx -y @hybridlabor-api/aos@latest --hooks-only`
See what would change first	`npx -y @hybridlabor-api/aos@latest --dry-run`
Check a broken setup	`aos doctor`
Open the local control page (services, catalog, store, plans)	`aos hub`
Browse and add skills and agents	`aos store ui`
Add the code knowledge graph	`npx -y @hybridlabor-api/aos@latest -y --modules=codebase-memory`
Plan a feature, then build it	`/plan`, then `/startcycle-graph`

See it in action

Skills at a glance

/startcycle - linear build pipeline (Architect, TechLead, parallel build, Reviewer) with file hand-offs in production_artifacts/
/startcycle-graph - dispatcher graph with durable state.json, Reviewer repair loop, quality gate and escalation
- The dispatcher invokes the subagents architect, techlead, godmode-ui-ux, godmode-engineering, godmode-media-eventtech, reviewer and godmode-shipping; they never call each other
/startcycle-graph-user - throwaway 2-4 node fan-out for any project, nothing persistent left behind
/plan - draft, render in a browser canvas, annotate, await approval, then hand off
- plan-canvas - local canvas where you annotate elements, chat and approve or request changes
- plan-arbiter - compare competing plans from several agents and produce one recommended plan
/bdbrainstorm - multi-agent brainstorm that ends in a hand-off to /startcycle-graph
- /grill-me - relentless interview to sharpen a plan or design
- bdbmediastorm - brainstorming for live event tech, show control and real-time media
/playbooks - list every pb-* playbook with duration, difficulty and requirements, and start one
- pb-bug-fix - GitHub issue to tested fix and PR; push and PR only after GO
- pb-ship - triage, review and gate the day's PRs; merge only after GO
- pb-release-aos - release AOS to npm through the release-please PR
- pb-show-build - build a show across grandMA3, Resolume and TouchDesigner, offline first
- pb-offer - draft a client offer priced from your price list; send only after GO
factory-collect - Factory: collect and triage feedback, telemetry, errors and issue reports
- factory-lookback - audit recurring problems across sources for systemic fixes
- factory-review-prs - review a configured PR queue; approve and merge stay with you
- factory-human-digest - read-only digest of what still needs a human decision
gogate - show or explain the go-gate mode (hard, soft, off) and the time-limited grants; CLI: aos-gogate
aos-a2a - list, send to and reply to live harness peers over A2A on localhost
- master-session - supervise several sessions: roster, status requests, GO board; workers can be spawned through aos-acp
- mcsc - delegate a task to another installed CLI harness (agy, OpenCode, Codex)
agenttrail - live browser map of a multi-agent build; CLI: aos-trail
aos-store - browse, preview and install skills and agents from a local web UI (aos store)
memb-skill - local-first long-term memory engine (memB)
- memb-ingest - ingest project files and conversation logs into memB
openwiki-skill - initialize, update and visualize codebase wikis with OpenWiki
synapse-integration-skill - BDB Synapse 3D codebase visualizer integration
/doctor - check the machine and the current project, print suggested permissions.allow entries, write no settings
aos-setup - bring a machine to a complete, verified AOS installation
- aos-project-init - set up a project folder: slug, wiki, memory, Synapse map, AGENTS.md
Godmodes - domain rulebooks
- godmode-engineering - strict DDD, TypeScript strictness, Clean Architecture, 5-step debugging triage
- godmode-ui-ux - brand discovery, Anti-Slop rules, DTCG design tokens, fluid motion
- godmode-eventtech - show-control execution for grandMA3, Resolume, Unreal, Rhino, Vectorworks and Adobe MCP
- godmode-hardware-pcb - schematics, PCB layout, KiCad and OpenSCAD enclosures with DFM/DRC/ERC sign-off
- godmode-shipping - final gatekeeper: pre-launch checks, feature-flag rollouts, rollback planning
Vendored helpers from BuilderIO/skills
- read-the-damn-docs - forces a docs pass before coding third-party APIs from memory
- stay-within-limits - respect 5-hour and weekly usage limits in long or parallel runs
- quick-recap - end each response with a red/yellow/green status line

Full catalogs (every skill, subagent, playbook and MCP server) are in the collapsible lists under What's included.

What's new in v5

v5 changes how AOS gets onto a machine and how its parts talk to each other.

Area	What changed
Installer	Rebuilt flow: Kernel, Harnesses, Packages, Optional, Verify. Profiles `minimal`, `standard`, `full`; `--no-hooks`, `--hooks-only`, `--without=<ids>`; repair and uninstall from the manifest; one version check at the start with Update as the default. See Install and docs/install-options.md.
MCP gateway	`aos-gateway`: one local endpoint (1mcp behind a token-checking forwarder) in front of an allow list of shipped servers, `aos-gateway adopt` for your own stdio servers, and one MCP per app (Blender, Resolume, After Effects). Installed with the default profile; opt out with `aos-gateway direct`, `--without=gateway` or `--profile=minimal`. docs/mcp-gateway.md.
A2A	Live Claude Code, OpenCode, agy and Codex sessions message each other with `aos-a2a`. An inbound message is never a GO. Any MCP-capable harness can join through the `aos-a2a` MCP server. docs/a2a.md.
Native agents	Subagents are compiled into the formats and paths each harness really reads, including OpenCode and agy. No model is pinned unless you set one: docs/agent-models.md.
Fleet mod	A band in Claude Code: gate mode, token-weather (context percent as a weather word plus a sparkline), and which sessions work or wait for you with their GO counts. `aos-gogate` shows gate state read-only.
AOS Hub	One local page for everything AOS runs on demand, on `127.0.0.1:7900`: Services, Catalog, Store and Plans. `aos hub`, `aos services`, `aos restart`. It replaces the old Launchpad HTTP server; `aos-dashboard` is now a thin alias. docs/hub.md.
agenttrail	One live map per repo, started automatically, with a link in your session. A v2 control room is in progress.
Plan Canvas	Plans open in a local canvas; annotations are dispatched back to the agent.
go-gate	Read, test and build commands with no write or network effect pass without a GO; deny logs mask credentials; hooks in four harnesses; a machine-wide gate profile (`strict`, `work`, `off`) narrows or widens it.
Playbooks and Factory	35 `pb-` playbooks, the nine `factory-` skills and `plan-arbiter`.
Orchestrator chain	master session, project orchestrator, package orchestrator, mapped to real mechanisms.

In progress, not claimed as shipped. agenttrail v2 control room, A2A intercom and a2abook, A2A spawn mode, skill-create and instinct-*.

The dispatcher graph

The graph is harness-neutral and runs on Claude Code's Dynamic Workflows, Antigravity's parallel execution, and any compatible agent harness.

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart TB
    subgraph SG[" "]
    direction TB
    U(["User"])
    A["<b>Architect</b><br/>System Plan"]
    T["<b>TechLead</b><br/>Capability Map"]
    UX["<b>UI_UX</b><br/>Frontend"]
    EN["<b>Engineering</b><br/>Backend"]
    ME["<b>Media_EventTech</b><br/>Creative"]
    R["<b>Reviewer</b><br/>QA"]
    S["<b>Shipping</b><br/>Gate"]

    U --> A --> T
    T --> UX & EN & ME
    UX & EN & ME --> R
    R --> S

    T -.->|reject| A
    R -.->|findings| UX
    S -.->|gate fail| EN
    R -.->|escalate| U
    end

    classDef plan fill:#E6F4EA,stroke:#1A7F3C,color:#15171A
    classDef build fill:#FFFFFF,stroke:#15171A,color:#15171A
    classDef gate fill:#1A7F3C,stroke:#1A7F3C,color:#FFFFFF
    classDef human fill:#F7F4EC,stroke:#5B6168,color:#15171A
    classDef store fill:#F7F4EC,stroke:#B9B5AA,color:#15171A
    linkStyle default stroke:#6e7681,stroke-width:2px

    style SG fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

    class A,T plan
    class UX,EN,ME build
    class R,S gate
    class U human

Install

Requirements. Node.js >= 20. macOS, Linux and Windows (PowerShell).

npx -y @hybridlabor-api/aos@latest

First run. The installer detects which harnesses are present, asks which to target and which tier (Pro MEDIA or Basic), copies the skills into each harness's skill directory, compiles the subagents, wires the hooks, merges the MCP configuration into each harness's own config file (existing entries are kept), and offers the optional modules listed below.

Every later run opens a menu instead:

Menu item	What it does
Quick Update	Refreshes skills, templates, hooks and installed modules to the version you just ran
Run System Checkup / Doctor	Runs `aos doctor`: dependencies, file placement, daemons, hooks
Drop Local Project Harness	Copies the dispatcher contract into the current directory (see below)
Reconfigure System	Change targets, tier, profile or options
Uninstall AOS	Removes what the installer placed; your data stays

Profiles and options

Flag	Effect
`--profile=standard` (default)	Skills, agents, rules, all hooks, MCPs, OpenWiki, fleet. Token Saver is off by default (use `--profile=full` to include it); for CLI output compression install rtk yourself, AOS does not install it.
`--profile=minimal`	Skills, agents and harness rules only. No context-injection hooks, OpenWiki, Token Saver, Codenotch, optional modules, gateway or fleet. The low-context choice.
`--profile=full`	`standard` plus every optional module (same as `--modules=all` with `-y`).
`--no-hooks`	Skips only the context-injection hooks (`memb-inject`, `rules-inject`, `trail-relay`, `trail-autostart`).
`--hooks-only`	Installs or refreshes hooks and their settings entries, then exits. With `--no-hooks` it refreshes the gates only.
`--without=id,id`	Excludes packages by registry id (`memb, synapse, openwiki, remote, ao, creator, hardware, rustauri, styleseed, installer, deja, token-saver, codenotch, gateway, hub, fleet, sessions, alert`) or the bundle `memory` (memB, deja and OpenWiki: unified memory). Saved, so updates and repair skip them.
`--modules=id,id`	Includes only the named optional modules on an otherwise non-interactive run (for example `--modules=creator` or `--modules=codebase-memory`); `all` takes every module. Also settable through `AOS_MODULES`.
`AOS_DISABLED_MCPS=a,b`	Keeps those shipped MCP server names out of the harness configs the installer writes.
`--dry-run`	Prints what would change and writes nothing.

The gate and safety hooks are always installed. No profile or flag removes go-gate, go-token, go-grant, graph-gate, conventional-commits or env-file-protection. The OpenCode plugin is not affected by --no-hooks or minimal. Profile choices are saved in ~/.aos/v5-settings.json. Full reference: docs/install-options.md.

Non-interactive

npx -y @hybridlabor-api/aos@latest -y --platforms=2          # Claude only, all defaults
npx -y @hybridlabor-api/aos@latest -y --platforms=1,5 --mcps=none
npx -y @hybridlabor-api/aos@latest -y --modules=codebase-memory,styleseed
npx -y @hybridlabor-api/aos@latest --dry-run                 # print what would change

--platforms= values: 0 universal (all detected), 1 Antigravity, 2 Claude Desktop / Claude Code, 3 Cursor, 5 Codex CLI, 6 Windsurf, 7 Roo Code / Cline, 8 Aider, 10 AOS CLI. 4 (custom paths) needs the interactive menu. --mcps=<name,name>|all|none picks the MCP subset. --verbose and --no-intro do what they say.

Local project harness

Instead of installing into $HOME, drop only the dispatcher contract (.agents/, the gate hooks, the agent definitions, the /startcycle-graph workflow, the OpenCode plugin) into one repository:

cd your-project && npx -y @hybridlabor-api/aos@latest --project-harness -y

Supported harnesses

What gets written. The installer writes the following for each target. Paths are the defaults; the installer only writes to harnesses it actually detects.

Harness	Skills	Subagents	Hooks	Plugin / rules
Claude Code / Claude Desktop	`~/.claude/skills`	`~/.claude/agents`	`~/.claude/hooks` + `settings.json` (GO gate, graph gate, env-file protection, Conventional Commits, memB inject, trail relay, A2A inbox)	`.claude-plugin/` manifest ships in the repo (see Contributing); fleet mod registered in `settings.json`
Google Antigravity	`~/.gemini/config/skills`	`~/.gemini/config/agents/<name>/agent.md`	`~/.gemini/config/hooks.json` and `~/.gemini/antigravity-cli/hooks.json`	root `plugin.json` / `plugins/bdb-aos/plugin.json`, commands `/bdb-aos:<cmd>`, see docs/codex-agy-setup.md
Codex CLI	`~/.agents/skills` (Codex reads this root and `~/.codex/skills`; AOS writes only the first, so skills are not listed twice)	`~/.codex/agents`	`~/.codex/hooks` + `config.toml`	`.codex-plugin/` + `.agents/plugins/marketplace.json`, commands `$bdb-aos:<cmd>`, see docs/codex-agy-setup.md
OpenCode	`~/.config/opencode/skills`	`~/.config/opencode/agents`	via plugin	`bdb-aos.js` plugin + `/startcycle-graph` command, registered in `opencode.jsonc`; keeps a `/startcycle-graph` run moving on `session.idle`; generated `/bdb-aos-<cmd>` commands; opt-in extras, see docs/opencode-setup.md
Cursor	`~/.cursor/skills`	none	none	`.cursor/rules` (project)
Windsurf	`~/.windsurf/bdb-skills`	none	none	`mcp.json`
Roo Code / Cline	`~/.roo/skills`	none	none	`.roomodes` (project)
Aider	`~/.aider/bdb-skills`	none	none	none
AOS CLI (`pi`)	reads `~/.agents/skills`	`~/.agents/AGENTS.md` as system prompt	none	no MCP; separate install, Node >= 22.19, see packages/aos-cli
BDB AO Codenotch (macOS app, Windows installer)	macOS `/Applications` or `~/Applications`; Windows per-user NSIS install (`/S`, no admin)	none	none	installed by default on macOS and Windows (never Linux); opt out with `--no-codenotch` or `AOS_CODENOTCH=0`; failures only warn; DMG from the public releases repo created by Tim, SHA-256 verified, ad-hoc signed with quarantine removed, see docs/codenotch.md

Every install also writes the universal copy to ~/.agents/skills, which is what the AOS CLI and the skills CLI read.

OpenCode agents. The four godmode personas (godmode-ui-ux, godmode-engineering, godmode-media-eventtech, godmode-shipping) are compiled with mode: all, so they show up in the OpenCode agent picker and can drive a whole session, not only a subagent call. Two thin session-level roles are written next to them as mode: primary agents that load the matching skill: master-session (roster, status requests, GO board) and project-orchestrator (parallel agents across repositories through the AO daemon). Every other agent is compiled as mode: subagent.

Capability tiers

The harnesses do not get the same safety layer. The tier says what is actually wired, not what a harness could do in theory; per-row evidence is in docs/harness-capabilities.md.

Tier	Harnesses	What you get
1 Full	Claude Code	GO gate as an enforced hook, typed grants and modes (`gogate grant ...`), Fleet panel, A2A inbox drained into the next prompt, all context hooks, MCP gateway entries
2 Gate and grants	OpenCode	GO gate through the plugin, grants capped at 2 h and no `off` mode, A2A injected as a prompt, MCP gateway entries; no panel
3 Gate hook only	Codex CLI, Antigravity (agy)	A `PreToolUse` gate hook is installed; grants and modes are not wired (only a literal human `GO` counts); Codex needs a one-time `/hooks` trust, and whether the hook fires under `codex exec`, `aos-acp` or `agy --print` is unverified; A2A inbound by `codex queue` (Codex) or pull-only (agy)
4 Skills, rules, MCP	Cursor, Windsurf, Roo Code / Cline, Aider	Skills and project rules (Cursor `.mdc`, Roo modes); MCP entries for Cursor; no hook, so the Destructive Actions text in `AGENTS.md` is the only safety layer
Not supported	Zed, Kimi	No installer target

Where a tier has no hook, nothing mechanical stops a guarded command: use a Tier 1 or 2 harness for anything that can push, publish or delete.

Delegation, A2A and gate hooks. Rows state what the code in this repo does. Source: docs/harness-capabilities.md.

Capability	Claude Code	OpenCode	Codex	agy
Delegation out (`aos-acp`)	yes	yes	yes	via shell commands (no ACP client of its own)
A2A inbound	UserPromptSubmit drain + Stop nudge	plugin inject as synthetic prompt	`codex queue` / `codex exec resume`	pull only, inbox MCP `a2a_inbox_pull`; a running interactive session cannot be pushed into
A2A reply	`aos-a2a reply` via pre-approved Bash rules	reply file written by the plugin	answer to the queued or resumed prompt	`a2a_reply` tool, or `aos-a2a reply`
A2A sidecar mode	`--mode live`	`--mode live`	`--mode live`	`--mode live`
GO gate hook	yes	yes (plugin, shared hooks)	yes (a hook is installed; firing under `codex exec`, `aos-acp` and `agy --print` is unverified)	gate-only hooks (`PreToolUse`, `PreInvocation`, `Stop`)
Never-GO on A2A inbound	yes	yes	yes	yes

The pipelines

The contract lives in .agents/graph.md, the node roster in .agents/nodes.json, the executable dispatcher in .claude/workflows/startcycle-dispatch.mjs.

One rule: nodes never invoke each other. A dispatcher reads production_artifacts/state.json after each node returns and decides what runs next. There is no hand-off chain and no agent telling another agent to go. This design ensures context fidelity, single-place auditability of routing logic, and portability across harnesses.

Each node reads the plan and its own prior state, executes its work, writes its artifact and state fragments, and returns. The dispatcher merges per-node state fragments (state.d/<node>.json), evaluates edge predicates, and routes to the next node, or escalates to the user if a no-progress guard triggers (same blocking finding on the second repair cycle) or the iteration ceiling is reached.

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart LR
    subgraph SG[" "]
    direction LR
    U(["User"]) --> A("Architect") --> T("TechLead")
    T --> UX("Godmode_UI_UX") & EN("Godmode_Engineering") & ME("Godmode_Media")
    UX & EN & ME --> R("Reviewer") --> S("Shipping")
    T -.->|reject| A
    R -.->|findings| UX
    R -.->|needs_human| U
    end

    classDef plan fill:#E6F4EA,stroke:#1A7F3C,color:#15171A
    classDef build fill:#FFFFFF,stroke:#15171A,color:#15171A
    classDef gate fill:#1A7F3C,stroke:#1A7F3C,color:#FFFFFF
    classDef human fill:#F7F4EC,stroke:#5B6168,color:#15171A
    classDef store fill:#F7F4EC,stroke:#B9B5AA,color:#15171A
    linkStyle default stroke:#6e7681,stroke-width:2px

    style SG fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

    class A,T plan
    class UX,EN,ME build
    class R,S gate
    class U human

Command	Machinery	Use when
`/startcycle`	Linear chain, file hand-offs in `production_artifacts/`. No state machine, no repair loop. `/startcycle --skill=<name> <goal>` forces a skill into every node.	A straightforward build with the agent roster but without the ceremony.
`/startcycle-graph`	The full graph: durable `state.json`, Reviewer repair loop, automated quality gate, human escalation.	Feature work where correctness matters more than speed and you want an audit trail.
`/startcycle-graph-user`	Throwaway 2-4 node fan-out. Nothing persistent. Model-tiered per role; uses Antigravity, OpenCode or Codex if installed, Claude Code subagents otherwise.	A one-off "spawn a few workers" in any project.

The three variants side by side

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart TB
    subgraph L["/startcycle (linear)"]
        direction LR
        L1("Architect") --> L2("TechLead") --> L3("parallel build") --> L4("Reviewer")
        L1 -. "production_artifacts/*.md" .-> L4
    end
    subgraph G["/startcycle-graph (durable)"]
        direction LR
        D{{"Dispatcher<br/>state.json"}}
        D --> N1("node") --> D
        D --> N2("node") --> D
        D --> Q("quality gate")
        D -. "no progress / ceiling" .-> H(["Human"])
    end
    subgraph F["/startcycle-graph-user (throwaway)"]
        direction LR
        F0("main session") --> W1("worker") & W2("worker") --> F3("review pass")
    end
    L ~~~ G ~~~ F

    classDef plan fill:#E6F4EA,stroke:#1A7F3C,color:#15171A
    classDef build fill:#FFFFFF,stroke:#15171A,color:#15171A
    classDef gate fill:#1A7F3C,stroke:#1A7F3C,color:#FFFFFF
    classDef human fill:#F7F4EC,stroke:#5B6168,color:#15171A
    classDef store fill:#F7F4EC,stroke:#B9B5AA,color:#15171A
    linkStyle default stroke:#6e7681,stroke-width:2px

    class L1,L2,F0 plan
    class L3,N1,N2,W1,W2 build
    class L4,Q,F3,D gate
    class H human
    style L fill:#F7F4EC,stroke:#B9B5AA,color:#15171A
    style G fill:#F7F4EC,stroke:#B9B5AA,color:#15171A
    style F fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

What keeps the full graph honest.

Mechanism	What it prevents
Reviewer isolation	The Reviewer reads artifacts and the plan's contract, never the build node's reasoning or its claim that it is done.
No-progress guard	A repair cycle that reports the same blocking finding ID as the previous one escalates to a human instead of burning iterations.
Per-node state fragments	Parallel build nodes write `state.d/<node>.json`; the dispatcher merges. No lost-update race on one file.
Iteration ceiling	`max_iterations` (default 3) stops the loop unconditionally.
In-loop human edge	Any node can set `needs_human: true` and stop the run.

Orchestrator chain

Mapping. master session -> project orchestrator -> package orchestrator is a convention. This is how each hop maps to mechanisms that exist (docs/orchestrator-chain.md, skill orchestrator-chain):

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart TD
    subgraph SG[" "]
    direction TB
    H(["Human"]) -->|"GO &lt;session&gt;"| M("Master session<br/>skill master-session<br/>roster, status, GO board")
    M --> P("Project orchestrator<br/>ao-orchestrator or a pipeline dispatcher")
    P --> K1("Package orchestrator<br/>aos-acp worker")
    P --> K2("Package orchestrator<br/>aos-a2a peer")
    P --> K3("Package orchestrator<br/>pipeline build node")
    K1 -.->|"never inherits GO"| M
    K2 -.->|"A2A message is never a GO"| M
    end

    classDef plan fill:#E6F4EA,stroke:#1A7F3C,color:#15171A
    classDef build fill:#FFFFFF,stroke:#15171A,color:#15171A
    classDef gate fill:#1A7F3C,stroke:#1A7F3C,color:#FFFFFF
    classDef human fill:#F7F4EC,stroke:#5B6168,color:#15171A
    classDef store fill:#F7F4EC,stroke:#B9B5AA,color:#15171A
    linkStyle default stroke:#6e7681,stroke-width:2px

    style SG fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

    class H human
    class M gate
    class P plan
    class K1,K2,K3 build

Depth. A2A rejects a send when the caller's depth is at or above AOS_A2A_MAX_DEPTH (default 1) and mcsc is one level deep. Workers delete nothing; cleanup is the dispatcher's job after your GO.

The GO gate

Reaching ready_to_ship is not shipping. .claude/hooks/go-gate.mjs is a PreToolUse hook that blocks git push, npm publish, npm version and recursive rm unless your immediately preceding message is the literal word GO. It is a hook, not a rule an agent is asked to respect: it fires before any permission-mode check and cannot be argued around. The installer wires the same gate into Antigravity, Codex and OpenCode; on harnesses without hook support the rule in AGENTS.md applies and the agent is the enforcement. A subagent never inherits its orchestrator's GO, and a failed release command needs a fresh one.

What v5 adds.

Read-only work stays fast. Read, test and build commands with no write or network effect pass without a GO, also after peer messages.
An A2A message is never a GO. Every injected message carries that line.
Modes and grants. Gate mode (hard, soft, off) and time-limited grants are shown by aos-gogate status and the gogate skill. aos-gogate is read-only and never records a grant.
Deny logs mask credentials (bearer tokens, flags, URLs with credentials).
Limit. The hook sees tool calls the harness routes through it. A worker started with permissions bypassed asks for nothing, so the in-harness hook stays the main layer.

Gate profiles

A machine-wide, human-set profile widens or narrows what the gate asks for. .claude/hooks/go-gate.mjs reads it through .claude/hooks/gate-profile.mjs, so every harness that uses the shared go-gate (Claude Code, OpenCode, Codex, agy) honours it.

Profile	Needs you	Default / maximum
`strict`	everything guarded (default)	never expires
`work`	the release class only	4 h / 24 h
`off`	hard blocks only	1 h / 2 h

aos-gate status                       # profile, time left, last change
aos-gate profile work --for 24h       # needs a real TTY; refused inside a harness session
aos-gate profile strict               # no confirmation needed

Release class (still needs a GO under work): push to the default branch, main/master or release/*, npm publish, npm version, npm dist-tag add/rm, gh pr merge, gh release create, recursive delete, bulk move out of home and delete outside the own worktree.
Hard blocks stay in every profile: writes to the gate store or its config, delete paths built from HOME, ~ or environment variables, recursive delete outside the own worktree, edits to the gate hooks and the settings that register them.
Fail closed: a missing, unreadable, symlinked, oversized, wrongly owned or wrong-mode (not 0600) config, a corrupt or expired file, or a work/off profile without expiry all read as strict. Config: ~/.aos/gate/config.json.
Loosening requires typing back a random word on the TTY.
Accepted residual: code running as your own user can write the profile file directly. See docs/gate-residuals.md.

Playbooks

A playbook is a skill with kind: playbook that turns a recurring job into a guided run with a fixed contract. The 35 pb-* playbooks ship with the skills; list and start them with /playbooks (it runs list-playbooks.mjs and prints name, time, difficulty and requirements for each). There is no per-playbook command: you name the playbook.

What a playbook declares. Its frontmatter names trigger phrases, inputs, requires (skills, subagents, MCP servers, store items), go_points, outputs, a verify check, difficulty and est_time. The body is a numbered step list. Each step names the skill or agent it uses, its input, the artifact it writes and the check that says it is done.

How a run works.

One session by default. The main session walks the steps and calls the required skills. Every step appends one line to a run log (production_artifacts/pb-<name>-<date>.md, never committed).
Subagents where it matters. Steps marked (agent) run a subagent: reviewer in pb-bug-fix, pb-ship and pb-release-aos; architect, techlead and reviewer in pb-harness-work, pb-idea-to-launch and pb-redesign-app; security-reviewer and silent-failure-hunter in pb-security-sweep; opensource-forker and opensource-sanitizer in pb-open-source. On a harness without subagents the main session reads the agent file and runs the same contract inline. pb-master goes one step further and runs a control session over several Claude Code, Codex or OpenCode sessions.
GO points are part of the contract. A [GO] step stops with WAITING FOR GO: <step> until you type GO; one GO covers one step, once. The go-gate hook is only the backstop. A failed check stops the run and is written into the log; nothing is retried silently.
One command, not one per playbook. There is a single playbooks command that lists them; after that you name the playbook. How to reach it differs per harness:

Harness	Type this
Claude Code	`/bdb-aos:playbooks`, then name one; or say what you want ("master session" starts `pb-master`)
Codex	`$bdb-aos:playbooks`; mentioning a playbook by name in words also works through `~/.agents/skills` (direct `$pb-<name>` use is unverified)
OpenCode	`/bdb-aos-playbooks`, or ask in words; playbooks that need an app server (media, show) stop with "Missing MCP" by design
Antigravity (agy)	`/bdb-aos:playbooks` (command namespacing on a live agy is unverified, see docs/agy-coverage.md)
Cursor, Windsurf, Roo Code / Cline, Aider	skills folder only, no command: ask for the playbook by name

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart LR
    subgraph SG[" "]
    direction LR
    I[("Issue")] --> C1("Root cause<br/>+ test")
    C1 --> C2("Fix")
    C2 --> R("Review")
    R -.->|findings| C2
    R --> C3("PR")
    C3 --> G{{"GO"}}
    G -->|yes| C4("Push")
    end

    classDef plan fill:#E6F4EA,stroke:#1A7F3C,color:#15171A
    classDef build fill:#FFFFFF,stroke:#15171A,color:#15171A
    classDef gate fill:#1A7F3C,stroke:#1A7F3C,color:#FFFFFF
    classDef store fill:#F7F4EC,stroke:#B9B5AA,color:#15171A
    linkStyle default stroke:#6e7681,stroke-width:2px

    style SG fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

    class I store
    class C1,C3 plan
    class C2,C4 build
    class R,G gate

Playbook	Purpose	Subagents	Time · level
Build and code
`pb-bug-fix`	Turn a GitHub issue into a tested fix on a branch with an open PR.	reviewer	30-90 min · intermediate
`pb-project-new`	Start a new private GitHub project the AOS way.	—	20-40 min · intermediate
`pb-idea-to-launch`	Turn an idea into a deployed prototype.	architect, techlead, reviewer	2-6 h · advanced
`pb-harness-work`	Change the agent harness itself (hooks, gates, memory, permissions, plugins) the safe way.	architect, techlead, reviewer	1-3 h · advanced
`pb-security-sweep`	Security sweep over one repo: secrets, dependencies, diff, one ranked findings report.	security-reviewer, silent-failure-hunter	1-2 h · advanced
`pb-open-source`	Prepare a project for open sourcing: sanitized fork, sanitizer verdict, README and LICENSE, new private repo.	opensource-forker, opensource-sanitizer	1-3 h · advanced
`pb-worktrees-land`	Clean up git worktrees across your repos; removes only merged ones.	—	10-30 min · intermediate
Release, CI and operations
`pb-ship`	Ship the day's work in one repo: triage, review every open PR, merge.	reviewer	20-60 min · intermediate
`pb-release-aos`	Release a new AOS version to npm through the release-please PR, merged only after GO.	reviewer	30-60 min · advanced
`pb-ci-fix`	Fix red CI or set up GitHub Actions end to end; push only after GO.	—	15-45 min · intermediate
`pb-deploy-saas`	Deploy a SaaS app to the BDB fleet: preflight, guardrail plan, green CI, deploy after GO, health check.	—	30-90 min · advanced
`pb-health-weekly`	Weekly health report over the BDB repos: version drift and CI status.	—	15-30 min · intermediate
Design and web
`pb-landing-page`	Build and launch a landing page: confirmed copy, brand tokens, UI and SEO review, deploy after GO.	—	1-3 h · intermediate
`pb-redesign-app`	Overhaul one app's UI against an audit.	architect, techlead, reviewer	2-4 h · advanced
`pb-docs-site`	Publish a project's docs as a static HTML manual on GitHub Pages.	—	30-90 min · intermediate
`pb-newsletter`	Turn a recap or changelog range into a newsletter draft where every claim cites a source line.	—	20-40 min · intermediate
Media and event tech
`pb-show-build`	Build a show across lights (grandMA3), media (Resolume) and visuals (TouchDesigner) from one cue list, offline first.	—	2-6 h · advanced
`pb-crew-call-sheet`	Build a crew call sheet and a load-in / load-out plan for one show day; send to the crew only after GO.	—	10-20 min · beginner
`pb-event-tracker`	One spreadsheet for an event: guests, vendors, timeline, budget, with vendor messages drafted.	—	20-40 min · intermediate
`pb-clip-from-moodboard`	Turn a look brief and reference images into a finished social clip.	—	1-3 h · advanced
`pb-image-to-3d`	Turn one reference image into a cleaned, scaled 3D asset.	—	30-90 min · advanced
`pb-launch-video`	Turn a live app or landing page into a short launch video.	—	30-90 min · intermediate
`pb-social-pack`	Turn a release recap into a social pack where every claim traces to a facts file.	—	30-60 min · intermediate
`pb-content-studio`	Idea to published-ready content: brief, script, render through the Creator Extension content-studio skill, then a social pack.	—	45-120 min · intermediate
Hardware
`pb-pcb-to-case`	From a KiCad board to a parametric OpenSCAD enclosure that fits it.	—	1-3 h · advanced
Management and office
`pb-offer`	Draft a client offer with line items priced only from your price list; send only after GO.	—	10-20 min · beginner
`pb-invoice-check`	Check invoices and receipts line by line against your offers. Nothing is paid or sent.	—	10-20 min · beginner
`pb-inbox-zero`	Sort an email backlog into reply, delegate, archive and ignore, and draft the replies.	—	10-20 min · beginner
`pb-meeting-actions`	Turn a meeting transcript into notes, decisions and owners; send the follow-up only after GO.	—	10-20 min · beginner
`pb-handover`	Write a handover note for a colleague from the state of a project folder.	—	10-20 min · beginner
`pb-week-plan`	Turn scattered to-do lists and notes into one prioritised plan for the week.	—	10-20 min · beginner
`pb-focus-chunks`	Split one big task into chunks of at most 25 minutes, each with a done-check.	—	5-10 min · beginner
`pb-todo`	Turn one sentence into a task line in the right to-do list.	—	2-5 min · beginner
Machine and sessions
`pb-machine-setup`	Bring a new machine to a verified AOS installation.	—	30-60 min · intermediate
`pb-master`	One control session over several Claude Code, Codex or OpenCode sessions: roster, status board, GO board.	—	15 min setup, then session-long · advanced

Factory

What it is. Nine experimental skills adapted from BuilderIO/skills that turn feedback, telemetry, errors, issues and pull requests into a repeatable review loop: factory (the schedule and policy that hosts the others), factory-collect, factory-lookback, factory-review-prs, factory-human-digest, plus the delivery-side factory-ship, factory-babysit-pr, factory-watchdog and factory-recover. In AOS they report and draft; every external write waits for your GO. plan-arbiter ships alongside them for comparing competing plans.

The review loop.

Skill	Reads	Writes	Needs GO?
`factory-collect`	the sources in `workflows.collect.sources`	triage report, draft replies, proposed close wording; optional local fix	Reply, close, push, merge, publish
`factory-lookback`	configured sources over a bounded period, prior fixes	pattern report; optional local systemic fix	Reply, close, publish, merge
`factory-review-prs`	PRs matching the configured filters	per-PR findings, draft comments, approval and merge readiness	Reply, approve, merge
`factory-human-digest`	configured repositories and sources (default last 7 days)	decision queue, nothing else	Read-only

How to run

Invoke by name. Ask the agent to run the skill, for example factory-collect; each SKILL.md carries a "Use when" trigger. Start by hand and read the report before enabling anything scheduled.
Configure. .agent-factory/config.yaml, read by all of them. Keys named in the skills: workflows.collect.sources, workflows.collect.implement, workflows.lookback, workflows.lookback.implement, workflows.human-digest, and an optional skill_prompts.<skill-name> per skill. If the file is missing the skill stops and asks; it never creates it.
The GO rule. Nothing in the config opens the go-gate. git push, gh pr merge, gh release create and every reply, comment, approval, close or status change need your literal GO for that exact action. Scheduled or unattended runs are read-only.

Vendored from BuilderIO/skills (MIT) with AOS safety changes, see THIRD_PARTY_NOTICES.md; upstream docs: https://github.com/BuilderIO/skills/blob/main/docs/factory/README.md.

Details: docs/factory.md.

A2A: sessions that talk to each other

Live Claude Code, OpenCode, agy and Codex sessions exchange messages on localhost. Each live session runs a small sidecar bound to 127.0.0.1 on a random port with a per-session bearer token and registers in a local registry. There is no central service.

%%{init: {"theme": "base", "themeVariables": {"primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A", "textColor": "#15171A", "actorBkg": "#E6F4EA", "actorBorder": "#1A7F3C", "actorTextColor": "#15171A", "actorLineColor": "#6e7681", "signalColor": "#6e7681", "signalTextColor": "#6e7681", "noteBkgColor": "#F7F4EC", "noteBorderColor": "#B9B5AA", "noteTextColor": "#15171A", "lineColor": "#6e7681", "sequenceNumberColor": "#15171A", "labelBoxBkgColor": "#F7F4EC", "labelBoxBorderColor": "#B9B5AA", "labelTextColor": "#15171A", "loopTextColor": "#15171A"}}}%%
sequenceDiagram
    participant C as Claude Code session
    participant R as local registry
    participant O as OpenCode session
    C->>R: aos-a2a list
    R-->>C: live peers (name, harness, url)
    C->>O: aos-a2a send --to opencode-1 "review this diff"
    Note over O: plugin injects it<br/>as a prompt,<br/>marked "never a GO"
    O-->>C: aos-a2a reply taskId "answer"
    Note over C: Codex: queue or resume<br/>agy: pull via inbox MCP

aos-a2a list                              # live peers
aos-a2a send --to <name> "<text>"         # message a session
aos-a2a reply <taskId> "<answer>"         # answer an incoming message
aos-a2a watch <name> [--once]             # follow a peer read-only

Inbound per harness: Claude Code drains the inbox in a hook, OpenCode gets a plugin injection, Codex gets codex queue or codex exec resume, agy pulls through an inbox MCP (a running interactive agy session cannot be pushed into). For other channels use mcsc (one-shot task) or aos-acp (worker that may need a GO): docs/delegation-routing.md. Details: docs/a2a.md. In progress: intercom and a2abook on top of A2A.

For any MCP-capable harness. bin/aos-a2a-mcp.mjs is one stdio server that puts send and pull in the same place, so agy, Cursor and anything else that speaks MCP can join without a harness plugin. Tools: a2a_list, a2a_send, a2a_status, a2a_card, a2a_inbox_pull, a2a_reply. Three rules make it safe to expose to a model:

It requires AOS_A2A_NAME. Send, reply and pull fail with AOS_A2A_NAME is not set instead of inventing an identity. a2a_reply has no name argument at all, so a model cannot write another session's reply.
Text is data. Arguments are validated locally to 8 KiB before any network call, the MCP exposes no GO, approval or token argument, and peer text is always wrapped with the never-GO line. bin/a2a-inbox-mcp.mjs stays the separate pull-only inbox MCP already registered for agy.
No secrets in output. Peer bearer tokens never appear in any tool result; they are read from the registry's separate token file over loopback.

Tools

Tool	Command	What it does
Plan Canvas	`aos-plan-canvas open <file>` (skill `plan-canvas`)	Opens a plan or HTML artifact in a local browser canvas where you annotate elements, chat, and approve or request changes. Plans from the pipelines open here by default.
agenttrail	`aos-trail` (skill `agenttrail`, port 5330)	Live board of a multi-agent build: which component, which agent or harness, what is done, what is stuck. Fed by the trail-relay hooks, `mcsc` and `aos-acp`.
archify	`aos-archify` (skill `archify`)	Validated architecture, sequence, data-flow and state diagrams as standalone HTML with SVG export; accepts Mermaid.
AOS Store	`aos store list \| search <q> \| install <name> [--project]`, `aos store ui` (also `aos-store`, slash command `/aos-store`)	Browse and install AOS Core, ECC and Scenario (scenario-labs/skills, MIT) skills and agents; required skills are installed together and every file is SHA-256 verified. Both UIs lay the items out as a card grid (installed vs missing, name, category, tier) instead of a table: the standalone web UI on `http://127.0.0.1:4322` and the Store tab of the AOS Hub, where a slide-over previews the exact target paths and installs or rolls back only after you type the name. `list` and `search` read a pinned offline index: 287 ECC skills (library 174, bdb-core 46, engineering-method 38, design-ui-ux 19, media-eventtech 10), 53 ECC subagents and 13 Scenario skills (third-party, MIT, scenario-labs/skills). Items the installer already places are excluded from the list.
AOS Hub	`aos hub [--port 7900] [--tab id] [--no-open]`, `aos hub status [--json]`, `aos hub stop`, slash command `/hub`	The one local page for everything AOS runs on demand, on `127.0.0.1:7900`, with four tabs: Services (status, start/stop/restart, log tails), Catalog (every skill, agent, command and MCP with what is installed where), Store (the card grid) and Plans (plan-kit sessions linked to their agenttrail map). Loopback only, per-process token in `~/.aos/hub/hub.json` (0600), typed confirm on anything destructive. It replaces the old Launchpad HTTP server; `aos-dashboard` is now a thin alias that starts the hub. A fresh install registers it as a login service (launchd, systemd user unit or Task Scheduler); `--without=hub` skips that.
Launchpad	`aos-dashboard [--port 7900] [--no-open]`	Thin alias for the AOS Hub on the same port. Superseded by `aos hub`.
Sessions	`aos-sessions [--json]`, Claude Code mod `bdb-aos-sessions` (`/aos-sessions`)	Read-only overview of every live session across harnesses plus the ACP workers, in four groups: NEEDS YOU, WORKING, IDLE, ACP WORKERS. Feeds the fleet band and the hub Plans tab. The installer enables the mod when Claude Code is found.
Doctor	`aos doctor [--json] [--net]` (also `aos-doctor`)	Verifies dependencies, skill placement per harness, daemons, hooks, modules and the MCP gateway; exits 1 when something needs attention. The first thing to run when anything misbehaves.
Workers	`aos-acp <claude\|codex\|opencode\|agy> --name <worker> [--sandbox coop\|wsl2]`	Start a worker whose guarded commands need a GO token for its name. `--sandbox` wraps the launch in an optional per-run sandbox (off by default): `coop` on macOS/Linux, `wsl2` on Windows.
Fleet band	Claude Code mod `plugins/bdb-aos-fleet`	Two-line band: gate mode (click it for the gate pane), token-weather, sessions that work or wait with GO counts, and the active projects. Needs Claude Code >= 2.1.287; enabled by the installer, opt out with `AOS_NO_FLEET=1` or `--without=fleet`.
Alert bar	Claude Code mod `plugins/bdb-aos-alert` (`/aos-alert test`, `/aos-alert off`)	A red blinking bar around the prompt when a go-gate block, permission request or question stops the session. Like the sessions mod it is read-only and grants no GO. Enabled by the installer like the fleet and sessions mods (`--without=alert` to skip).
GO helper	`aos-gogate status [--session <id>] \| preset <name> \| presets`	Read-only gate status and preset texts. Never records a grant.
Gate profile	`aos-gate status \| profile <strict\|work\|off> [--for <duration>]`	Read or change the machine-wide gate profile; changing one needs a real TTY and is refused inside a harness session.
A2A	`aos-a2a list \| send \| reply \| status \| cancel`	Messages between live harness sessions.
MCP gateway	`aos-gateway status \| enable \| direct \| adopt`	One local endpoint for shared MCP servers. Installed by a fresh install with the default profile; `direct` turns it back off.
Config	`aos-config show \| propose \| set <key> <value>`	Machine-level `~/.agents/aos-config.json`: workspace root, domains, user id.
mcsc	MCP tools `delegate_agy`, `delegate_opencode`, `delegate_codex`, `delegate_smart` (skill `mcsc`)	Delegates a task to another installed CLI harness and streams its tool calls to agenttrail. Preferred over shelling out to the CLI.

AOS Hub and the sandbox

Hub tabs. Services, Catalog, Store and Plans. The Catalog tab is strictly read-only: it reads skills/, agents/, commands/, mcp_config.json, mcp_picks.json and the gateway allow list, and cross-checks the command list against plugin.json. No request value ever reaches the filesystem and nothing is written. The same CLI works without a browser: aos services status [--json], aos services start <id>, aos services stop <id> --confirm <id>, aos restart <id> --confirm <id>. Details: docs/hub.md.

Worker sandbox. aos-acp --sandbox <provider> wraps one worker launch, and only that launch. It is off by default and the provider is checked before anything starts:

Provider	Where	How it wraps
`coop`	macOS and Linux	Needs the `coop` binary on `PATH` and `AOS_SANDBOX_COOP_CMD`, a template containing `{cmd}` as its own token (for example `coop <subcommand> {cwd} -- {cmd}`), because the pinned coop v0.8.0 README documents no run-a-command form. `{cwd}` is replaced with the worker directory.
`wsl2`	Windows	`wsl.exe --cd <path> --exec <cmd...>`, so the prompt is not re-split by a shell. Drive paths are translated to `/mnt/<drive>/...`; UNC and `\\?\` paths are refused rather than mistranslated. Unverified on a real Windows machine.

An unknown provider, or one unavailable on this platform, fails with the reason and nothing is launched. --sandbox does not narrow paths itself: with --allow-default=scoped the path narrowing is agy's sandbox and the aos-go-gate hook, not this flag.

Fleet band and token-weather

Token-weather. The fleet mod turns the context percent of a session into a weather word with advice and a sparkline: for example ok at low use, compact soon around 80 percent, compact or start a new session above 90. The same band lists your Claude sessions, which work, which wait for you, and how many GO requests each has pending. Source: plugins/bdb-aos-fleet.

MCP servers

mcp_config.json defines 21 local stdio servers, built or warmed by the installer from mcps/. They reach your harnesses in one of two ways:

Direct. Each server is written into every harness's MCP configuration, so every harness sees the same tool set. This is what you get with --without=gateway, --profile=minimal or after aos-gateway direct.
Through the MCP gateway (the default). Every harness gets one aos entry pointing at 127.0.0.1:7790; behind it, 1mcp runs the servers on the gateway allow list once for all harnesses. aos-gateway adopt moves your own local servers behind it too, and aos-gateway direct switches back. deja, memB, mcsc and the OS-control servers always stay direct.

One residual worth knowing. The gateway entry carries its local token inline as an X-AOS-Token header in each harness config: the secret launcher (bin/aos-mcp-env.mjs) covers stdio servers, and a streamable-HTTP entry needs the header itself. The token stops browsers, not other local processes.

The column "Routing" says which path a server takes when the gateway is on; "Reaches the app via" says how it talks to its application. API keys and tokens live in ~/.aos/secrets.env (mode 0600), never inline in a harness config: docs/mcp-secrets.md.

Server	App	Reaches the app via	Routing	Skill
3D / CAD
`bdb_blender_mcp`	Blender	socket integration; the official Blender MCP is picked instead when Blender >= 5.1 is found	gateway (official pick: direct)	`bdb-blender-mcp`
`bdb_rhino_mcp`	Rhino 3D, Grasshopper	McNeel Yak router, needs Rhino on the host	direct	`bdb-rhino-mcp`
`bdb_rhino_mcp_fallback`	Rhino 3D	GOLEM 3D (105 tools)	direct	`bdb-rhino-mcp`
`bdb_unreal_mcp`	Unreal Engine 5	Web Remote Control API, port 30010	gateway	`bdb-unreal-mcp`
Video / post
`bdb_davinci_mcp`	DaVinci Resolve	Resolve scripting API (162 tools)	gateway	`bdb-davinci-mcp`
`bdb_after_effects_mcp`	After Effects	pinned `@kumoproductions/mcp-aftereffects`, Node >= 24; legacy server only with `AOS_AE_MCP=legacy`	gateway	`bdb-after-effects-mcp`
`bdb_after_effects_mcp_fallback`	After Effects	Go server, ExtendScript	direct	`bdb-after-effects-mcp`
`adobe_uxp_mcp`	Photoshop, Illustrator, Premiere Pro, After Effects	UXP WebSocket bridge	gateway	`bdb-adobe-suite-mcp`
Show control / live
`bdb_grandma3_mcp`	grandMA3	typed `ma3_*` tools, OSC/UDP port 8000	gateway	`bdb-grandma3-mcp`
`bdb_resolume_mcp`	Resolume Arena	REST API, port 8080; the official Arena server is picked when its binary is found	gateway (official pick: direct)	`bdb-resolume-mcp`
`bdb_td_minddesigner`	TouchDesigner	MindDesigner bridge, port 9980	direct	`bdb-touchdesigner-mcp`
`bdb_td_backup`	TouchDesigner	stdio bridge, fallback	direct	`bdb-touchdesigner-mcp`
Design / browser
`open_design_mcp`	Open Design	local daemon on 127.0.0.1:3000	gateway	—
`chrome-devtools`	Chrome	Puppeteer	direct	—
OS control
`zavora_computer_use`	macOS, Linux desktop	`npx -y @zavora-ai/computer-use-mcp@7.4.0`, native Rust module bundled in the npm package	direct	`bdb-computer-use-mcp`
`bdb_windows_computer_use`	Windows desktop	Win32, COM, UIAutomation, local Tesseract OCR	direct	`bdb-computer-use-mcp`
Memory / infra
`memb_mcp`	memB	local SQLite + ONNX, WebUI on port 8088	direct	`memb-skill`, `bdb-memb-mcp`
`deja`	agent transcripts	local index, secrets redacted	direct	`deja-memory`
`mcsc`	other harnesses	spawns agy, OpenCode or Codex and streams to agenttrail	direct	`mcsc`
`github`	GitHub	`@modelcontextprotocol/server-github`	gateway	`github`
`bdb_remoteos_mcp`	RemoteOS	multi-cloud gateway with 4-eyes approval	direct	—

Routing. The allow list in lib/gateway/config.js decides: the nine servers marked gateway are shared through the one aos entry once you enable it; everything else stays direct in every harness.

MCP gateway and one MCP per app

One MCP per app. For Blender, Resolume and After Effects the installer keeps a single candidate (mcp_picks.json): Blender official when Blender >= 5.1 is found, Resolume official when the Arena binary is found, the bundled server otherwise. The After Effects server is a pinned npx package and needs Node >= 24 (the legacy server only with AOS_AE_MCP=legacy).

Gateway (the default profile). aos-gateway runs a token-checking forwarder on 127.0.0.1:7790 in front of @1mcp/agent and shares an allow list of shipped servers behind one aos entry per harness. aos-gateway adopt moves your own local stdio servers behind it (skipping remote entries and entries with inline secrets). aos-gateway direct reverses everything. The token stops browsers, not other local processes. deja, memb_mcp and mcsc stay direct. Opt out with --without=gateway or --profile=minimal.

Skipped shapes. The gateway only writes into a harness whose config shape it has verified. Claude Desktop, Cline, Aider, VS Code, Windsurf, the Cursor app-support copy and a second agy candidate are reported as skipped: shape unverified and are never modified, because no vendor documentation for their remote-entry shape is recorded. The verified rows are Claude Code, Cursor (~/.cursor/mcp.json), Roo Code, OpenCode, Codex (config.toml) and Antigravity.

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart LR
    subgraph D["Direct (gateway off)"]
        direction LR
        H1("Claude") --> S1[("github")] & S2[("blender")] & S3[("resolume")]
        H2("OpenCode") --> S1 & S2 & S3
        H3("Codex") --> S1 & S2 & S3
        H4("agy") --> S1 & S2 & S3
    end
    subgraph G["Gateway (default)"]
        direction LR
        G1("Claude") & G2("OpenCode") & G3("Codex") & G4("agy") --> F("aos entry<br/>127.0.0.1:7790<br/>token check")
        F --> M("1mcp agent<br/>127.0.0.1:7791") --> U1[("github")] & U2[("blender")] & U3[("resolume")]
        G1 -.->|"stays direct"| X[("memb_mcp, deja, mcsc")]
    end
    D ~~~ G

    classDef plan fill:#E6F4EA,stroke:#1A7F3C,color:#15171A
    classDef build fill:#FFFFFF,stroke:#15171A,color:#15171A
    classDef gate fill:#1A7F3C,stroke:#1A7F3C,color:#FFFFFF
    classDef human fill:#F7F4EC,stroke:#5B6168,color:#15171A
    classDef store fill:#F7F4EC,stroke:#B9B5AA,color:#15171A
    linkStyle default stroke:#6e7681,stroke-width:2px

    class H1,H2,H3,H4,G1,G2,G3,G4 plan
    class S1,S2,S3,U1,U2,U3,X store
    class F,M gate
    style D fill:#F7F4EC,stroke:#B9B5AA,color:#15171A
    style G fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

Reference. docs/mcp-gateway.md.

Plugins and marketplace

Plugin manifest. .claude-plugin/plugin.json + marketplace.json (generated by npm run plugin:build). AOS ships plugin manifests for Claude Code (.claude-plugin/), Codex (.codex-plugin/) and Antigravity (plugin.json); OpenCode has no marketplace concept and loads a local plugin instead. Sources: docs/marketplaces.md.

Harness	Install	Update
Claude Code	`claude plugin marketplace add hybridlabor-api/aos` · `claude plugin install bdb-aos@bdb-marketplace`	`claude plugin marketplace update bdb-marketplace` · `claude plugin update bdb-aos@bdb-marketplace`
Codex CLI	`codex plugin marketplace add hybridlabor-api/aos` · `codex plugin add bdb-aos@bdb-aos`	`codex plugin marketplace upgrade bdb-aos` · `codex plugin add bdb-aos@bdb-aos`
Antigravity (agy)	`agy plugin install <local plugin dir>` — the npm installer runs this for you	re-run the npm installer
OpenCode	local plugin `.opencode/plugins/bdb-aos.js`, copied by the installer; no marketplace	overwrite the plugin file

The Claude marketplace bdb-marketplace offers four plugins: bdb-aos (skills and subagents), bdb-aos-fleet, bdb-aos-sessions and bdb-aos-alert. When Claude Code is found, the installer offers the three mods (-y enables them) and registers them in ~/.claude/settings.json itself; --without=fleet,sessions,alert or --profile=minimal skips them, and Uninstall removes them. The npm installer still performs each harness's plugin registration best-effort; the CLI commands above are the manual equivalents. The marketplace route carries skills and subagents only — the npm installer is the recommended route because it also sets up gate hooks and MCP servers.

Skills discovery. Every harness finds skills in its native directory (~/.claude/skills, ~/.agents/skills, ~/.codex/skills, ~/.roo/skills, etc.). To browse and install additional skills after install:

npx skills add hybridlabor-api/aos

This discovers all 270 curated skills and installs them into the universal ~/.agents/skills directory (used by all harnesses and the AOS CLI).

AOS CLI

A lightweight CLI harness built on pi, a coding agent that runs in the terminal. AOS CLI reads ~/.agents/skills (written by the installer) and ~/.agents/AGENTS.md (the dispatcher graph as system instructions), runs no MCP servers of its own, and needs Node >= 22.19 (pi's floor, higher than the main AOS installer).

aos-cli "what is the fastest way to fix this bug"
aos-cli --continue                    # resume the previous session

The CLI launcher (packages/aos-cli/bin/aos-cli.mjs) ships with an AOS-themed dark mode (aos.json), the ten core skills from core-skills.json (ask-tim, aos-setup, systematic-debugging, archify, etc.), and two in-session read-only commands (/aos shows the install menu; /aos-status runs the health check).

Install it through the AOS installer with the AOS CLI target: npx -y @hybridlabor-api/aos@latest -y --platforms=10. The package is private and is not published on npm, so npm i -g @hybridlabor-api/aos-cli does not work.

Memory and knowledge

Installed as optional modules by the installer; aos doctor verifies them and the AOS Hub Services tab shows them.

memB (@hybridlabor-api/memb): local, offline vector memory with an MCP server (add_memory, search_memory, list_memories, delete_memory), a WebUI on port 8088, and an ambient hook that injects relevant memories into Claude Code sessions. Skills: memb-skill, memb-ingest, bdb-memb-mcp.
deja (@vshulcz/deja-vu, installed with memB): indexes your agent transcripts locally with secrets redacted; deja fix on an error, deja wip when resuming, deja search for past sessions. Skill: deja-memory.
OpenWiki (openwiki CLI): generates and refreshes a grounded wiki of a codebase, with a visualizer on port 4321 and a background daemon. Skill: openwiki-skill; this repo's own wiki is under .openwiki/.
Synapse (@hybridlabor-api/bdb-synapse): renders a repository as a 3D code city and replays agent sessions through it. Skill: synapse-integration-skill.

aos-setup brings a machine to a verified state for all four; aos-project-init binds one project to them (slug, wiki, memory, AGENTS.md).

What's included

The subagents (45)

Roster. 45 agents are compiled into the native agent format of Claude Code, Antigravity, Codex and OpenCode; Cursor, Windsurf, Roo and Aider get none and use the rules in AGENTS.md instead. Sources are in agents/. AOS pins no model: agents inherit the session model unless .aos/pipeline.json overrides a role per harness (docs/agent-models.md).

Group	Agents
Pipeline nodes (7)	`architect`, `techlead`, `godmode-ui-ux`, `godmode-engineering`, `godmode-media-eventtech`, `reviewer`, `godmode-shipping`
Review and safety (6)	`security-reviewer`, `silent-failure-hunter`, `database-reviewer`, `reviewer-1`, `reviewer-2`, `devils-advocate`
Language reviewers (8)	`typescript-reviewer`, `python-reviewer`, `go-reviewer`, `rust-reviewer`, `java-reviewer`, `kotlin-reviewer`, `cpp-reviewer`, `csharp-reviewer` — read-only, compiled with edit and bash denied
Build helpers (3)	`go-build-resolver`, `opensource-forker`, `opensource-sanitizer`
Domain assistants (5)	`spec-prd-creator`, `pcb-design-assistant`, `event-manager-assistant`, `live-media-operator`, `video-motion-assistant`

What the seven pipeline nodes do. Architect turns the goal into a system plan after reading the existing architecture. TechLead turns the plan into a capability map (module boundaries, dependency direction, build order) before any build node starts, and rejects back to Architect if it does not hold. godmode-ui-ux leads the frontend, godmode-engineering the backend, godmode-media-eventtech the creative and show-control work. Reviewer is an adversarial correctness review of build-node output against the plan's contract; Shipping runs the mechanical gate (lint, typecheck, tests, a11y, seo) and enforces the GO gate. The auxiliary 22 are not nodes of the graph. Subagents never inherit a GO and never invoke each other.

Skills by category

Categories. 270 curated skills, discoverable by every harness (the full generated catalog is in the next section). Eight categories, enforced by npm run validate:

bdb-core: Core AOS infrastructure, pipelines, tools, and utilities: startcycle, startcycle-graph, startcycle-graph-user, agenttrail, plan-canvas, aos-a2a, aos-gateway, aos-store, master-session, codebase-memory, and more.
design-ui-ux: Frontend, UI design, accessibility, tokens, motion, anti-slop: senior-frontend, ui-ux-pro-max, tailwind-patterns, shadcn, wcag-audit-patterns, and more.
engineering-method: Architecture, testing, debugging, CI/CD, code quality: software-architecture, test-driven-development, systematic-debugging, ci-pipeline, github-actions-generator, docker-expert, makefile-pro, and more.
library: Language and framework specifics: TypeScript, Node.js, Python, React, Postgres, Prisma, Next.js, Drizzle ORM, Go, and more.
media-eventtech: 3D, video, show control, spatial design: godmode-eventtech, threejs-skills, the creative-software MCP skills, and more.
engineering-hardware: PCB and electrical design: godmode-hardware-pcb.
saas-ops: Running a SaaS end to end: golive, pb-deploy-saas, pb-health-weekly.
business-ops: The commercial office side: pb-offer, pb-invoice-check.

Browse everything

Generated catalog. Built from the repo files (frontmatter of every SKILL.md, agents/, commands/). Open a section to see names and one-line descriptions.

Harness	Skills	Subagents	Hooks and gates	MCP	Extras
Claude Code / Desktop	yes	yes	yes (GO gate, graph gate, env-file protection, Conventional Commits, memB inject, trail relay, A2A inbox)	yes	slash commands, fleet band, plugin manifest
Google Antigravity (agy)	yes	yes	gate-only hooks	yes	`/bdb-aos:<cmd>` commands, inbox MCP for A2A
Codex CLI	yes	yes	GO gate, graph gate	yes	`$bdb-aos:<cmd>` commands, A2A queue/resume
OpenCode	yes	yes	via `bdb-aos.js` plugin	yes	`/bdb-aos-<cmd>` commands, A2A inject
Cursor	yes	no	no	yes	`.cursor/rules`
Windsurf	yes	no	no	yes	`mcp.json`
Roo Code / Cline	yes	no	no	yes	`.roomodes`
Aider	yes	no	no	no	none
AOS CLI (pi)	yes	system prompt	no	no	10 core skills, dark theme

Command	What it does
`aos`	Installer menu: update, doctor, reconfigure, uninstall.
`aos doctor`	Verify dependencies, skill placement, daemons, hooks, modules, gateway.
`aos store`	Browse and install skills and agents (CLI and web UI on port 4322).
`aos-plan-canvas`	Open plans and HTML artifacts in the annotation canvas (port 4519).
`aos-trail`	agenttrail live map (port 5330).
`aos-archify`	Architecture and sequence diagrams as standalone HTML.
`aos-hub`	The AOS Hub on port 7900: Services, Catalog, Store, Plans. `aos-dashboard` is a thin alias.
`aos-services` (via `aos services`)	Status, start, stop and restart for every local service.
`aos-a2a`	List, send to and reply to live harness peers.
`aos-acp`	Start workers whose guarded commands need a GO token; `--sandbox coop\|wsl2` wraps one launch.
`aos-gateway`	Local MCP gateway: serve, enable, direct, adopt, status.
`aos-gogate`	Read-only gate status and presets.
`aos-gate`	Read or change the machine-wide gate profile.
`aos-sessions`	Read-only overview of every live session and ACP worker.
`aos-config`	Machine-level AOS configuration.

Server	Domain
`github`, `chrome-devtools`	Issues, PRs, workflows; browser automation and debugging
`bdb_unreal_mcp`	Unreal Engine (Web Remote Control, port 30010)
`bdb_rhino_mcp`, `bdb_rhino_mcp_fallback`	Rhino 3D and Grasshopper
`bdb_davinci_mcp`	DaVinci Resolve
`bdb_grandma3_mcp`	grandMA3 (typed `ma3_*` tools)
`bdb_resolume_mcp`	Resolume Arena (REST, port 8080)
`adobe_uxp_mcp`	Adobe UXP bridge
`bdb_blender_mcp`	Blender
`bdb_after_effects_mcp`, `bdb_after_effects_mcp_fallback`	After Effects
`bdb_td_minddesigner`, `bdb_td_backup`	TouchDesigner (port 9980)
`zavora_computer_use`, `bdb_windows_computer_use`	Desktop control on macOS and Linux, and on Windows
`memb_mcp`, `deja`	Local memory and transcript search
`mcsc`	Delegation to other harnesses
`open_design_mcp`	Open Design
`bdb_remoteos_mcp`	RemoteOS multi-cloud gateway with 4-eyes approval

Skill	What it does
`ai-agent-development`	AI agent development workflow for building autonomous agents, multi-agent systems, and agent orchestration with CrewAI, LangGraph, and custom agents.
`ai-product`	Every product will be AI-powered.
`api-design-principles`	Master REST and GraphQL API design principles to build intuitive, scalable, and maintainable APIs that delight developers and stand the test of time.
`api-patterns`	API design principles and decision-making.
`apify-lead-generation`	Scrape leads from multiple platforms using Apify Actors.
`apify-ultimate-scraper`	AI-driven data extraction from 55+ Actors across all major platforms.
`bash-linux`	Bash/Linux terminal patterns.
`brainstorming`	Use before creative or constructive work (features, architecture, behavior).
`browser-automation`	Browser automation powers web testing, scraping, and AI agent interactions.
`cloudflare-workers-expert`	Expert in Cloudflare Workers and the Edge Computing ecosystem.
`copywriting`	Write rigorous, conversion-focused marketing copy for landing pages and emails.
`crewai`	Expert in CrewAI, the leading role-based multi-agent framework.
`database-design`	Database design principles and decision-making.
`debugger`	Debugging specialist for errors, test failures, and unexpected behavior.
`deep-research`	Run autonomous research tasks that plan, search, read, and synthesize information into comprehensive reports.
`docker-expert`	Advanced Docker containerization: optimization, security hardening, multi-stage builds, orchestration.
`documentation`	Documentation generation workflow covering API docs, architecture docs, README files, code comments, and technical writing.
`drizzle-orm-expert`	Expert in Drizzle ORM for TypeScript: schema design, relational queries, migrations, and serverless database integration.
`firecrawl`	Search, scrape, and interact with the web via the Firecrawl CLI.
`firecrawl-sdk`	Integrate Firecrawl into product code: SDK setup and FIRECRAWL_API_KEY onboarding, /scrape, /search and /interact endpoints.
`gemini-api-dev`	The Gemini API provides access to Google's most advanced AI models.
`gemini-api-integration`	Use when integrating Google Gemini API into projects.
`geo-fundamentals`	Generative Engine Optimization for AI search engines (ChatGPT, Claude, Perplexity).
`git-advanced-workflows`	Master advanced Git techniques to maintain clean history, collaborate effectively, and recover from any situation.
`git-pr-review`	Generate a concise and structured PR description from commit history with minimal token usage.
`github`	Use the `gh` CLI for issues, pull requests, Actions runs, and GitHub API queries.
`github-actions-templates`	Production-ready GitHub Actions workflow patterns for testing, building, and deploying applications.
`github-workflow-automation`	Patterns for automating GitHub workflows with AI assistance.
`go-concurrency-patterns`	Master Go concurrency with goroutines, channels, sync primitives, and context.
`go-playwright`	Robust browser automation using Playwright Go.
`golang-pro`	Master Go 1.21+ with modern patterns, advanced concurrency, performance optimization, and production-ready microservices.
`landing-page-generator`	Generates high-converting Next.js/React landing pages with Tailwind CSS.
`linear-claude-skill`	Manage Linear issues, projects, and teams.
`llm-app-patterns`	Production-ready patterns for building LLM applications.
`llm-application-dev-ai-assistant`	Creating intelligent conversational interfaces, chatbots, and AI-powered applications.
`llm-prompt-optimizer`	Use when improving prompts for any LLM.
`llm-structured-output`	Get reliable JSON, enums, and typed objects from LLMs using response_format, tool_use, and schema-constrained decoding.
`local-llm-expert`	Master local LLM inference, model selection, VRAM optimization, and local deployment using Ollama, llama.cpp, vLLM, and LM Studio.
`microservices-patterns`	Microservices architecture patterns: service boundaries, inter-service communication, data management, resilience.
`modern-javascript-patterns`	Modern JavaScript (ES6+) features, functional programming patterns, and best practices.
`monorepo-management`	Build efficient, scalable monorepos that enable code sharing, consistent tooling, and atomic changes.
`n8n-code-javascript`	Write JavaScript code in n8n Code nodes.
`n8n-code-python`	Write Python code in n8n Code nodes.
`n8n-expression-syntax`	Validate n8n expression syntax and fix common errors.
`n8n-mcp-tools-expert`	Expert guide for using n8n-mcp MCP tools effectively.
`n8n-workflow-patterns`	Proven architectural patterns for building n8n workflows.
`neon-postgres`	Expert patterns for Neon serverless Postgres, branching, connection pooling, and Prisma/Drizzle integration.
`nextjs-app-router-patterns`	Next.js 14+ App Router architecture, Server Components, and modern full-stack React development.
`nextjs-best-practices`	Next.js App Router principles.
`obsidian-markdown`	Create and edit Obsidian Flavored Markdown with wikilinks, embeds, callouts, and properties.
`openapi-spec-generation`	Generate and maintain OpenAPI 3.1 specifications from code, design-first specs, and validation patterns.
`os-scripting`	Operating system and shell scripting troubleshooting workflow for Linux, macOS, and Windows.
`playwright-skill`	General-purpose browser automation skill.
`posix-shell-pro`	Expert in strict POSIX sh scripting for maximum portability across Unix-like systems.
`postgres-best-practices`	Postgres performance optimization and best practices from Supabase.
`postgresql`	Design a PostgreSQL-specific schema.
`prisma-expert`	Prisma ORM: schema design, migrations, query optimization, relations modeling, and database operations.
`product-manager-toolkit`	Essential tools and frameworks for modern product management, from discovery to delivery.
`programmatic-seo`	Design and evaluate programmatic SEO strategies for creating SEO-driven pages at scale.
`python-patterns`	Python development principles and decision-making.
`python-performance-optimization`	Profile and optimize Python code using cProfile, memory profilers, and performance best practices.
`python-pro`	Master Python 3.12+ with modern features, async programming, performance optimization, and production-ready practices.
`rag-engineer`	Expert in building Retrieval-Augmented Generation systems.
`rag-implementation`	RAG implementation workflow: embedding selection, vector database setup, chunking strategies, retrieval optimization.
`react-best-practices`	Performance optimization guide for React and Next.js applications, maintained by Vercel.
`react-component-performance`	Diagnose slow React components and suggest targeted performance fixes.
`react-patterns`	Modern React patterns and principles.
`readme`	Technical writer for comprehensive project documentation.
`remotion`	Generate walkthrough videos from Stitch projects using Remotion with transitions, zooming, and text overlays.
`schema-markup`	Design, validate, and optimize schema.org structured data for eligibility, correctness, and measurable SEO impact.
`seo`	Run a broad SEO audit across technical SEO, on-page SEO, schema, sitemaps, content quality, AI search readiness, and GEO.
`seo-audit`	Diagnose and audit SEO issues affecting crawlability, indexation, rankings, and organic performance.
`seo-technical`	Audit technical SEO across crawlability, indexability, security, URLs, mobile, Core Web Vitals, and structured data.
`tanstack-query-expert`	Expert in TanStack Query (React Query), asynchronous state management.
`tmux`	Expert tmux session, window, and pane management for terminal multiplexing and persistent remote workflows.
`turborepo-caching`	Configure Turborepo for efficient monorepo builds with local and remote caching.
`typescript-pro`	Master TypeScript with advanced types, generics, and strict type safety.
`using-neon`	Neon serverless Postgres: autoscaling, branching, instant restore, scale-to-zero.
`vector-database-engineer`	Expert in vector databases, embedding strategies, and semantic search implementation.
`vercel-ai-sdk-expert`	Expert in the Vercel AI SDK.
`vercel-deployment`	Expert knowledge for deploying to Vercel with Next.js.
`web-artifacts-builder`	Build powerful frontend claude.ai artifacts.
`web-performance-optimization`	Optimize loading speed, Core Web Vitals, bundle size, caching strategies, and runtime performance.
`web-scraper`	Multi-strategy intelligent web scraping.
`zustand-store-ts`	Create Zustand stores following established patterns with proper TypeScript types and middleware.

Skill	What it does
`agent-tool-builder`	Tools are how AI agents interact with the world: schema to error handling.
`agentic-harness-patterns`	Harness patterns for coding agents: memory, permissions, context engineering, delegation, skills, hooks, bootstrap.
`archify`	Validated architecture, workflow, sequence, data-flow, and state diagrams as explorable standalone HTML with inline SVG.
`architect-review`	Master software architect specializing in modern architecture.
`bdb-computer-use-mcp`	Native Rust/Node and Python computer-use servers to control macOS, Windows, and Linux desktops.
`bdb-security-audit`	Security auditing, diff analysis, defensive checklists, and vulnerability testing.
`bdb-shipping-skill`	Problem-framing pre-flight, one-way/two-way door classification, ADR-lite decision logging, post-ship outcome loop.
`ci-pipeline`	Set up a complete CI/CD pipeline for a project.
`clean-code`	The principles of "Clean Code".
`concise-planning`	Generate a clear, actionable, atomic checklist for a coding task.
`diagnosing-bugs`	Diagnosis loop for hard bugs and performance regressions.
`dispatching-parallel-agents`	Use when facing 2+ independent tasks that can be worked on without shared state.
`domain-modeling`	Build and sharpen a project's domain model.
`executing-plans`	Execute a written implementation plan in a separate session with review checkpoints.
`factory`	>-
`factory-babysit-pr`	>-
`factory-collect`	Experimental workflow for collecting and triaging product feedback, telemetry, runtime errors, and issue reports.
`factory-lookback`	Experimental workflow for auditing recurring feedback, telemetry, and errors to find systemic fixes.
`factory-recover`	>-
`factory-review-prs`	Experimental workflow for reviewing configured repositories' pull requests.
`factory-ship`	>-
`factory-watchdog`	>-
`finishing-a-development-branch`	Decide how to integrate finished work once implementation is complete and tests pass.
`github-actions-generator`	Create, generate, or scaffold GitHub Actions workflows and CI/CD pipelines.
`github-actions-validator`	Validate, lint, audit, fix GitHub Actions workflows.
`github-repo`	Standards and workflows for writing, sanitizing, and publishing high-quality GitHub repositories.
`godmode-engineering`	Senior Fullstack and Backend Engineer.
`godmode-shipping`	Release Gatekeeper, QA and Verification Auditor.
`grill-me`	A relentless interview to sharpen a plan or design.
`grill-with-docs`	Like grill-me, and builds the project's domain model (glossary and ADRs) as it goes.
`grilling`	Grill the user relentlessly about a plan, decision, or idea.
`makefile-pro`	Generate, scaffold, validate and lint Makefiles and .mk files: .PHONY targets, build automation, common mistakes.
`planning-with-files`	Use persistent markdown files as working memory on disk.
`pr-recap`	Visual recap page for a PR or git range: changed-files tree, per-file notes, Verified vs Not verified table, risks.
`prompt-engineer`	Transforms prompts into optimized prompts using frameworks (RTF, RISEN, Chain of Thought, and more).
`prompt-engineering-patterns`	Advanced prompt engineering techniques to maximize LLM performance, reliability, and controllability.
`prototype`	Build a throwaway prototype to answer a design question.
`read-the-damn-docs`	Read the docs before implementing, integrating, upgrading, or debugging anything third-party.
`requesting-code-review`	Verify work meets requirements after completing tasks or before merging.
`simplify-code`	Review a diff for clarity and safe simplifications, then optionally apply low-risk fixes.
`software-architecture`	Guide for quality-focused software architecture.
`subagent-driven-development`	Execute implementation plans with independent tasks in the current session.
`systematic-debugging`	Use on any bug, test failure, or unexpected behavior, before proposing fixes.
`tdd`	Test-driven development.
`test-driven-development`	Use when implementing any feature or bugfix, before writing implementation code.
`to-spec`	Turn the current conversation into a spec and publish it to the project issue tracker: no interview, just synthesis of what you've already discussed.
`triage`	Move issues and external PRs through a state machine of triage roles and write agent-ready briefs.
`two-axis-code-review`	Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes: Standards (does the code follow this repo's documented coding sta...
`using-git-worktrees`	Ensure an isolated workspace for feature work and plan execution.
`verification-before-completion`	Run verification commands and confirm output before claiming work is complete or passing.
`wayfinder`	Plan a huge chunk of work (more than one agent session can hold) as a shared map of decision tickets on your issue tracker, and resolve them one at a time un...
`wcag-audit-patterns`	Audit web content against WCAG 2.2 with actionable remediation strategies.
`webapp-testing`	Test local web applications with native Python Playwright scripts.
`writing-plans`	Write a plan for a multi-step task from a spec, before touching code.

Skill	What it does
`agent-manager-skill`	Manage multiple local CLI agents via tmux sessions with cron-friendly scheduling.
`agent-memory-mcp`	Hybrid memory system: persistent, searchable knowledge management for AI agents.
`agent-pipeline`	Reference for the seven-node dispatcher graph that /startcycle-graph runs.
`agenttrail`	Live map of a multi-agent build in the browser: which plan component, which agent or harness, what is done, what is stuck.
`ao-orchestrator`	Multi-project orchestrator across repositories using the Agent Orchestrator (AO) daemon.
`aos-a2a`	List, send to and reply to live harness peers (Claude Code, OpenCode, Codex, agy) over A2A on localhost.
`aos-gateway`	Use the local MCP gateway and the per-app MCP picks.
`aos-project-init`	Interview a project folder into AOS: slug and domain, OpenWiki, memB, Synapse map, AGENTS.md.
`aos-setup`	Bring a machine to a complete, verified AOS installation.
`aos-store`	Browse, preview, and install AOS Core, ECC and Scenario skills and agents from a local web UI.
`ask-tim`	Ask which skill or flow fits your situation.
`bdb-aos`	Entry point for the BDB Agent OS suite installed as a Claude Code plugin.
`bdb-deploy`	Build and deploy a project to a real server via rsync over SSH.
`bdb-memb-mcp`	Model Context Protocol interface to the memB persistent agent memory layer.
`bdb-updater`	Proactively check for and install updates to the AOS package via npm.
`bdbhtmlmanueldocs`	Design system, standalone HTML template and GitHub Pages hosting workflow for neutral developer documentation.
`bdbrainstorm`	Multi-agent brainstorming, /grill-me and the Core Godmodes, ending in a hand-off to /startcycle-graph.
`bdbresilience`	CI/CD error recovery, file-based locking, diagnostic triage, and two-phase GO gate resilience for multi-agent pipelines.
`codebase-memory`	>-
`deja-memory`	Use `deja fix` on an error, `deja wip` when resuming, `deja search` for past sessions.
`design-control-loop`	Interview the user to design an agentic control loop tailored to their codebase, then build it.
`factory-human-digest`	Experimental workflow for summarizing work that still needs human judgment.
`gogate`	Show or explain the AOS go-gate mode (hard, soft, off) and the time-limited grants of this session.
`instinct-export`	>-
`instinct-import`	>-
`instinct-status`	>-
`loop-templates`	Repeat a task on a schedule or until a condition holds (CI until green, follow a PR to merge).
`master-session`	One Claude Code session supervises several others: roster, status requests, GO board, idle notices, GO-token protocol.
`mcsc`	Delegate a task to another installed CLI harness (agy, OpenCode, Codex) from inside an AOS repo.
`memb-ingest`	Deep scan and ingest project files and past conversation logs into the local memB vector memory engine.
`memb-skill`	BDB local-first long-term memory engine (memB).
`openwiki-skill`	Initialize, update, and visualize codebase or personal knowledge wikis using OpenWiki.
`orchestrator-chain`	Hand work down the chain master session, project orchestrator, package orchestrator.
`plan-arbiter`	Compare, cross-review, merge, or arbitrate competing plans from multiple agents.
`plan-canvas`	Open plans and HTML artifacts in a local browser canvas where the human annotates, chats, and approves.
`quick-recap`	End each agent response with a red/yellow/green status line.
`skill-create`	>-
`startcycle`	Linear multi-agent build pipeline with file hand-offs in production_artifacts/.
`startcycle-graph`	The autonomous multi-agent build pipeline with durable state, repair loop and escalation.
`startcycle-graph-user`	A small, throwaway multi-agent fan-out in any project.
`stay-within-limits`	Respect 5-hour and weekly usage limits by checking usage between waves and pausing near the cap.
`subagent-setup`	Configure and synchronize multi-harness subagents with optional per-harness model overrides.
`synapse-integration-skill`	BDB Synapse (3D Codebase Visualizer) integration.
`teamwork-preview`	Interactive 9-step prompt crafting and delegation protocol for autonomous multi-agent teams.
`token-saver-config`	Context window output compression engine for CLI commands.

Skill	What it does
`bdb-adobe-suite-mcp`	Automate Photoshop, Illustrator, Premiere Pro, and After Effects via ExtendScript and UXP bridges.
`bdb-after-effects-mcp`	Compositions, layers, masks, keyframe animations, and ExtendScript in After Effects.
`bdb-blender-mcp`	Build 3D assets, apply materials, inspect scenes, and script bpy in Blender.
`bdb-davinci-mcp`	DaVinci Resolve timeline editing, media analysis, color grading, Fusion and Fairlight scripting.
`bdb-eventagency-skill`	Event agency operations: client intake, scoping, vendors, crew, production planning, pre-show logistics, wrap.
`bdb-grandma3-mcp`	Patch fixtures, execute console commands, and trigger macros on grandMA3.
`bdb-resolume-mcp`	Trigger clips, clear layers, adjust speeds, and query composition status in Resolume Arena.
`bdb-rhino-mcp`	Create and manipulate 3D models and run Grasshopper graphs in Rhino.
`bdb-touchdesigner-mcp`	TOP/CHOP chains, operator scripting, parameter inspection, and node-network debugging in TouchDesigner.
`bdb-unreal-mcp`	Control UE5 through a native C++ Automation Bridge plugin.
`bdb-vectorworks-mcp`	Search and retrieve Vectorworks Python and VectorScript API documentation.
`bdbmediastorm`	Brainstorm live event technology, show control, and real-time media systems.
`brag`	Turn the current project website into a short, shareable launch video.
`godmode-3d-creation`	3D meshes, text-to-CAD, and scene reconstruction via TRELLIS, TripoSR, or Text-to-CAD engines.
`godmode-eventtech`	Real-time performance and multimedia operator work: signal flows, OSC/DMX, MCP orchestration, hardware limits.
`godmode-media-creation`	Assemble video timelines, sync beats, and create media via OpenMontage, Palmier Pro, or TouchDesigner.
`mcp-manage`	Check capabilities and guide use of specialized MCP servers like Unreal, Rhino, DaVinci, or TouchDesigner.
`spline-3d-integration`	Interactive 3D scenes from Spline.design in web projects.
`threejs-skills`	Create 3D scenes, interactive experiences, and visual effects using Three.js.

Skill	What it does
`bdb-visual-edit`	Edit source from an element the human points at in a running local dev app.
`bdbdesignpro`	Animation engine choice, motion tokens, scroll effects, micro-interactions, motion accessibility.
`design-spells`	Curated micro-interactions, delightful animations, and subtle design details.
`editable-design`	Fixed-canvas editable visual designs: posters, marketing graphics, covers, menus, banners, social cards.
`frontend-dev-guidelines`	Senior frontend engineering under strict architectural and performance standards.
`godmode-ui-ux`	Lead Frontend Designer and UI Engineer.
`live-preview-canvas`	Local HTML mock with a built-in feedback layer for visual comments.
`senior-frontend`	React components, Next.js performance, accessibility, frontend code quality.
`shadcn`	Add, customize, and troubleshoot shadcn/ui components.
`tailwind-patterns`	Tailwind CSS v4, CSS-first configuration, container queries, design token architecture.
`ui-ux-pro-max`	Comprehensive design guide for web and mobile applications.
`ux-persuasion-engineer`	Behavioral UX: choice architecture, friction audits, commitment design, without coercion.

Skill	What it does
`godmode-hardware-pcb`	Electrical schematics, PCB layouts, KiCad projects, OpenSCAD enclosures: DFM/DRC/ERC sign-off and co-design.

Skill	What it does
`golive`	Take an agent-written app from repo to live production on the user's OWN accounts, with providers they choose (hosting, database, auth, payments, email, doma...

Playbook	What it does
`pb-bug-fix`	Turn a GitHub issue into a tested fix on a branch with an open PR.
`pb-ci-fix`	Fix red CI or set up GitHub Actions end to end; push only after GO.
`pb-clip-from-moodboard`	Turn a look brief and reference images into a finished social clip.
`pb-content-studio`	>-
`pb-crew-call-sheet`	Build a crew call sheet and a load-in / load-out plan for one show day.
`pb-deploy-saas`	Deploy a SaaS app to the BDB fleet: preflight, guardrail plan, green CI, deploy after GO, health check.
`pb-docs-site`	Publish a project's docs as a static HTML manual on GitHub Pages.
`pb-event-tracker`	One spreadsheet for an event: guests, vendors, timeline, budget.
`pb-focus-chunks`	Split one big task into chunks of at most 25 minutes, each with a done-check.
`pb-handover`	Write a handover note for a colleague from the state of a project folder.
`pb-harness-work`	Change the agent harness itself (hooks, gates, memory, permissions, plugins) the safe way.
`pb-health-weekly`	Weekly health report over the BDB repos: version drift and CI status.
`pb-idea-to-launch`	Turn an idea into a deployed prototype.
`pb-image-to-3d`	Turn one reference image into a cleaned, scaled 3D asset.
`pb-inbox-zero`	Sort an email backlog into reply, delegate, archive and ignore, and draft the replies.
`pb-invoice-check`	Check invoices and receipts line by line against your offers.
`pb-landing-page`	Build and launch a landing page: confirmed copy, brand tokens, UI and SEO review, deploy after GO.
`pb-launch-video`	Turn a live app or landing page into a short launch video.
`pb-machine-setup`	Bring a new machine to a verified AOS installation.
`pb-master`	One control session over several Claude Code, Codex or OpenCode sessions: roster, status board, GO board.
`pb-meeting-actions`	Turn a meeting transcript into notes, decisions and owners; send the follow-up only after GO.
`pb-newsletter`	Turn a recap or changelog range into a newsletter draft where every claim cites a source line.
`pb-offer`	Draft a client offer with line items priced only from your price list.
`pb-open-source`	Prepare a project for open sourcing: sanitized fork, sanitizer verdict, README and LICENSE, new private repo.
`pb-pcb-to-case`	From a KiCad board to a parametric enclosure that fits it.
`pb-project-new`	Start a new private GitHub project the AOS way.
`pb-redesign-app`	Overhaul one app's UI against an audit.
`pb-release-aos`	Release a new AOS version to npm through the release-please PR, merged only after GO.
`pb-security-sweep`	Security sweep over one repo: secrets, dependencies, diff, one ranked findings report.
`pb-ship`	Ship the day's work in one repo: triage, review every open PR, merge.
`pb-show-build`	Build a show across lights (grandMA3), media (Resolume) and visuals (TouchDesigner) from one cue list.
`pb-social-pack`	Turn a release recap into a social pack where every claim traces to a facts file.
`pb-todo`	Turn one sentence into a task line in the right to-do list.
`pb-week-plan`	Turn scattered to-do lists and notes into one prioritised plan for the week.
`pb-worktrees-land`	Clean up git worktrees across your repos; removes only merged ones.

Agent	What it does
`architect`	Turns the user's goal into a system plan.
`cpp-reviewer`	Independent read-only C++ code reviewer: memory safety, modern C++ idioms, concurrency, and performance for `.cpp`/`.hpp`/`.cc`/`.h` changes.
`csharp-reviewer`	Independent read-only C#/.NET code reviewer: .NET conventions, async patterns, security, nullable reference types, EF Core.
`database-reviewer`	PostgreSQL specialist for query optimization, schema design, security, and performance.
`devils-advocate`	Argues against the plan or diff under review: hunts unstated assumptions, missing failure modes, and the case where the proposed approach is the wrong one.
`event-manager-assistant`	Event production planning assistant: run of show, technical riders, crew plans, schedules and budgets, load-in planning, and vendor lists.
`go-build-resolver`	Resolves Go build, vet, and compilation errors with minimal changes.
`go-reviewer`	Independent read-only Go code reviewer: idiomatic Go, concurrency, error handling, and performance for `.go` changes.
`godmode-engineering`	Senior Fullstack and Backend Engineer.
`godmode-media-eventtech`	Creative-Tech and Show-Control Specialist.
`godmode-shipping`	Release Gatekeeper, QA and Verification Auditor.
`godmode-ui-ux`	Lead Frontend Designer and UI Engineer.
`java-reviewer`	Independent read-only Java code reviewer for Spring Boot and Quarkus projects: layered architecture, JPA/Panache, security, concurrency, testing.
`kotlin-reviewer`	Independent read-only Kotlin/Android/KMP code reviewer: idiomatic Kotlin, coroutine and Flow safety, Compose recomposition traps, clean architecture, Android...
`live-media-operator`	Live show operator for grandMA3, media servers such as Resolume, and TouchDesigner networks.
`opensource-forker`	Forks a project for open-sourcing: strips secrets and internal references.
`opensource-sanitizer`	Verifies an open-source fork is fully sanitized before release.
`pcb-design-assistant`	PCB design assistant for KiCad-centered hardware work: schematic and layout review, DRC/ERC clearance, stackup and impedance constraints, BOM checks, and man...
`python-reviewer`	Independent read-only Python code reviewer: Pythonic idioms, PEP 8, type hints, security, and concurrency for `.py` changes.
`reviewer`	Adversarial review of build-node output against the plan's contract.
`reviewer-1`	Independent adversarial reviewer, correctness focus: logic errors, broken edge cases, mismatches between the plan's stated contract and the implementation, a...
`reviewer-2`	Independent adversarial reviewer, security and operability focus: injection, leaked secrets, auth gaps, silent failures, and rollback/deployment hazards.
`rust-reviewer`	Independent read-only Rust code reviewer: ownership, lifetimes, unsafe usage, error handling, and idiomatic patterns for `.rs` changes.
`security-reviewer`	Security vulnerability detection and remediation.
`silent-failure-hunter`	Reviews code for silent failures, swallowed errors, bad fallbacks.
`spec-prd-creator`	Turns a raw idea into a spec/PRD: goals, non-goals, user stories, acceptance criteria, and open questions.
`techlead`	Reviews the plan for a capability map before any build node starts.
`typescript-reviewer`	Independent read-only TypeScript/JavaScript code reviewer: type safety, async correctness, Node/web security, and idiomatic patterns for `.ts`/`.tsx`/`.js`/`...
`video-motion-assistant`	Code-driven video and motion specialist, carved out of godmode-media-eventtech: programmatic video generation and editing, After Effects and Premiere automat...

Command	What it does
`/brainstorm`	Brainstorm a topic with the matching skill: bdbrainstorm, bdbmediastorm, grill-me.
`/doctor`	Check the machine and project, and print suggested permissions.allow entries without writing settings.
`/graph`	Run the startcycle-graph pipeline.
`/hub`	Open the AOS Hub on 127.0.0.1:7900 (services, plan kits, store) or stop/status it.
`/init`	Set up the current project folder for AOS.
`/loop`	Run a prompt repeatedly on an interval.
`/mastersession`	Supervise other Claude Code sessions: roster, status, GO board, idle notices.
`/memb`	Query or store long-term memory with memB.
`/orchestrator`	Coordinate parallel coding agents across repositories through the AO daemon.
`/plan`	Plan end to end: draft, render in plan-canvas, annotate, await approval, hand off.
`/playbooks`	List all pb-* playbooks and start one.
`/setup`	Bring this machine to a complete, verified AOS installation.
`/shipping`	Pre-ship framing and the technical release gate, with approvals under the go-gate rules.
`/startproject`	Start a new composition: project init, brainstorm, then graph or startcycle.
`/store`	Browse and preview AOS skills, agents and playbooks in the local store UI.

Skills

Persona layer. The Godmode skills are specialized personas that directly map to the build and ship nodes of the dispatcher graph.

All UI/UX work — the godmode-ui-ux persona and the design skills it routes through, such as bdbdesignpro — draws components first from the pinned BDB component and motion-icon library (@bdb shadcn registry backed by npm @hybridlabor-api/ui-motion-library), before falling back to third-party registries or building from scratch.

Godmode	Owns	Maps to
`godmode-engineering`	Domain-Driven Design, Clean Architecture, strict TypeScript/Python, systematic debugging, database best practices.	Engineering node
`godmode-ui-ux`	Anti-slop frontend principles, DTCG design tokens, motion dynamics, accessibility (WCAG), high-agency taste.	UI_UX node
`godmode-shipping`	Pre-launch checks, automated quality gates, safe rollback procedures, Go-gate enforcement.	Shipping node
`godmode-eventtech`	Show control, signal flow, DMX lighting, TouchDesigner networks, Resolume media servers, live-event hardware.	Media_EventTech node
`godmode-3d-creation`	MCP-first 3D generation, mesh reconstruction, parametric CAD, spatial modeling.	Optional specialist
`godmode-media-creation`	Video production, timeline assembly, motion design pipelines, OpenMontage, Remotion.	Optional specialist
`godmode-hardware-pcb`	Electrical schematics, PCB layout and routing, KiCad ERC/DRC/DFM gate, enclosure co-design, OpenSCAD.	Optional specialist

Entry points and navigation.

ask-tim: Skill recommendation by description
bdbrainstorm and bdbmediastorm: Multi-agent ideation sessions ending in an executable plan
teamwork-preview: Prompt crafting, role delegation, collaboration setup
Grilling family: grill-me (general audit), grill-with-docs (documentation-grounded), triage (prioritization)
CI/CD and generators: ci-pipeline, github-actions-generator, docker-expert, makefile-pro
Code quality: bdb-security-audit, systematic-debugging, silent-failure-hunter, bdbresilience
Framework specialists: Full coverage of TypeScript, React, Next.js, Drizzle ORM, Prisma, Python, Go, and more

Skills CLI. The library is also readable by the skills CLI:

npx skills add hybridlabor-api/aos

Playbooks

35 pb-* playbooks turn recurring jobs into guided runs with declared GO points. See Playbooks for how a run works and the full table by domain.

Optional modules

Optional by design. The installer's module picker offers these and Quick Update keeps them current. AOS works standalone without any of them. Pick them non-interactively with -y --modules=<id>,<id> or AOS_MODULES; the ids are the names below plus rustauri, styleseed and codebase-memory. Anything vendored ships inside the kernel, so those install with no download.

memB: local vector memory

@hybridlabor-api/memb: offline, local vector memory with an MCP server, WebUI on port 8088, and an ambient hook that injects relevant memories into Claude Code sessions. Skill: memb-skill, memb-ingest, bdb-memb-mcp.

memB is not an optional module any more: it is part of the default unified memory (memB, deja and OpenWiki, plus Codebase Memory MCP when you code), installed with the standard profile and with -y. Opt out with --without=memory or --profile=minimal; without it, no harness gets memb_mcp, deja or the memb-inject hook.

deja: transcript indexing

@vshulcz/deja-vu: indexes your agent transcripts locally (secrets redacted), with deja fix on an error, deja wip to resume, deja search for past sessions. Installed with memB. Skill: deja-memory.

OpenWiki: living documentation

openwiki CLI: generates and refreshes a grounded wiki of a codebase, with a visualizer on port 4321 and a background daemon. Skill: openwiki-skill. This repo's wiki: .openwiki/.

Synapse: @hybridlabor-api/bdb-synapse renders a repository as a 3D code city and replays agent sessions as light trails. Skill: synapse-integration-skill.

AO (Agent Orchestrator): @hybridlabor-api/bdb-agent-orchestrator orchestrates parallel agents in Git worktrees with live terminal control and CI/CD feedback. Skill: ao-orchestrator.

Creator Extension: @hybridlabor-api/bdb-dev-creator-extension provides ComfyUI MCP (FLUX, SDXL), image-to-3D (TripoSR, TRELLIS), and automated video production (OpenMontage, Remotion). Skill: bdb-dev-creator-extension.

Hardware and PCB: @hybridlabor-api/bdb-hardware-pcb drives KiCad and OpenSCAD design with ERC/DRC gate and parametric enclosure design. Skills: godmode-hardware-pcb, bdb-hardware-pcb.

Rust + Tauri (`rustauri`)

Ships inside the kernel under vendor/rustauri, nothing to download. Two skills: tauri-pilot, which drives and tests a running Tauri v2 app through the CLI or MCP (MIT, adapted from mpiton/tauri-pilot), and rust-pro, written for this repo. Enable with -y --modules=rustauri or aos package install rustauri.

The tauri-pilot MCP entry is registered in every harness even when the tauri-pilot CLI could not be installed (no cargo, or the build failed). The MCP server then fails at launch until the CLI is present; the install warning names the failed step.

StyleSeed (`styleseed`)

Ships inside the kernel under vendor/styleseed, nothing to download. Eight UI/UX skills in the StyleSeed Toss design language (MIT, from bitjaru/styleseed): ui-component, ui-page, ui-pattern, ui-review, ui-tokens, ux-audit, ux-feedback, ux-flow. It is opt-in because the skills assume that design system; the agent personas, playbooks and docs refer to them only as "if the StyleSeed package is installed". Copies from earlier installs are kept, not deleted. Enable with -y --modules=styleseed.

Codebase Memory MCP (`codebase-memory`)

The optional code knowledge graph. A third-party self-contained stdio binary (DeusData/codebase-memory-mcp, MIT) that indexes a repository into a local SQLite graph under ~/.cache/codebase-memory-mcp/ and answers structural questions ("who calls X", "what routes does service A expose") with far fewer tokens than reading files one by one. No cloud calls, no API keys.

AOS never downloads the binary. If you install it yourself (pinned to 0.11.0 when AOS installs it for you), the MCP entry is registered with the next harness MCP sync; aos doctor reports app missing: codebase-memory-mcp while it is missing. Asking for it is the "yes, I code here" answer: the installer asks "Do you use AOS for coding?" once, stores the answer as codingMode in ~/.aos/v5-settings.json, and --coding / --no-coding set it without a prompt. Enable with -y --modules=codebase-memory; details and the measurement recipe in docs/codebase-memory.md. Skill: codebase-memory.

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart LR
    subgraph S["Synapse"]
    direction LR
    A[("Logs")] -->|parse| B("Adapters")
    B --> C("Events")
    C & D[("Repo")] --> E("Go server")
    E --> F("WebGL")
    end
    style S fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart LR
    subgraph A["AO"]
    direction LR
    H(["IDE"]) --> O("Orchestrator")
    O --> S1("Build") & S2("Refactor")
    S1 & S2 --> C("Monitor")
    C --> F("Feedback")
    end
    style A fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart LR
    subgraph C["Creator"]
    direction LR
    A("Agent") -->|MCP| R("Router")
    R --> D("3D") & V("Video")
    D --> T["TRELLIS"]
    V --> O["OpenMontage"]
    end
    style C fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

%%{init: {"theme": "base", "themeVariables": {"lineColor": "#6e7681", "edgeLabelBackground": "#f6f1e8", "textColor": "#15171A", "clusterBkg": "#F7F4EC", "clusterBorder": "#B9B5AA", "titleColor": "#15171A", "primaryColor": "#FFFFFF", "primaryBorderColor": "#15171A", "primaryTextColor": "#15171A"}}}%%
flowchart LR
    subgraph P["Hardware/PCB"]
    direction LR
    A("Agent") -->|MCP| K("KiCad") & O("OpenSCAD")
    K --> E("ERC") & L("Layout")
    O --> EN("Enclosure")
    E & L & EN --> F[("Output")]
    end
    style P fill:#F7F4EC,stroke:#B9B5AA,color:#15171A

Heimdall Token Saver: CLI output compression

@hybridlabor-api/heimdall-token-saver: compresses repeated CLI output via ambient hooks on every harness. Reduces token overhead on large projects. Skill: token-saver-config.

Token optimization at a glance

Lever	How
Minimal profile	`--profile=minimal` skips context-injection hooks and optional modules.
No context hooks	`--no-hooks` keeps the gates and drops memB, rules and trail injection.
Heimdall Token Saver	Compresses repeated CLI output via hooks.
No pinned model	Agents inherit the session model; per-role overrides live in `~/.aos/pipeline.json` (docs/agent-models.md).
Token-weather	The fleet band shows how full the context is and when to compact.

Updating

Update. Run the same command again. The installer sees the installed version, offers Quick Update, and refreshes skills, hooks, templates and modules:

npx -y @hybridlabor-api/aos@latest

aos update. The installer puts an aos launcher in ~/.local/bin that runs npx -y @hybridlabor-api/aos@<channel> (latest, or beta when you installed a beta), so aos update, aos status and aos repair always run the newest release of your channel. If you once ran npm i -g @hybridlabor-api/aos, whichever aos comes first on your PATH wins: a global npm copy stays frozen at its version, so update it (npm i -g @hybridlabor-api/aos@latest) or remove it. The installer prints the update command itself whenever a newer version exists; the bdb-updater skill wraps the same check for use from inside a session.

Uninstall

aos-uninstall              # removes what AOS installed; memory, wikis and credentials stay
aos-uninstall --purge      # also removes ~/.MemBDB, ~/.openwiki, ~/.synapse, ~/.memb
aos-uninstall --dry-run    # list everything, delete nothing

How it works. The uninstaller works from the install manifest: a file that still matches the hash AOS wrote is removed, a file you edited is backed up instead, a file AOS never wrote is not touched. The same action is in the installer menu. aos-uninstall --restore-plugin-backup restores the loose skill copies the installer removed when it registered the plugin; see docs/plugin-migration.md. Uninstall also moves your saved install options to ~/.aos/v5-settings.json.removed-<timestamp> instead of deleting them.

FAQ

Do I need all nine harnesses? No. The installer only writes to harnesses it detects.

Will it overwrite my MCP config? No. Existing entries are kept and AOS merges its own. AOS_DISABLED_MCPS keeps chosen servers out.

Can I install with minimal context? Yes: --profile=minimal, or --no-hooks. The gate and safety hooks stay either way.

Does the GO gate work on every harness? It is a real hook on Claude Code, Antigravity, Codex and the OpenCode plugin. On Cursor, Windsurf, Roo / Cline and Aider only the rule in AGENTS.md applies.

Can an agent approve its own push? Not through the gate: only your own typed GO counts, not an agent message, a subagent's relay or an A2A message.

Does AOS work without AO, memB or OpenWiki? Yes. All of them are optional modules.

Something is broken. Run aos doctor. It names what is missing and exits 1.

Contributing

AGENTS.md is the single source of rules for every harness: the skill contract, category routing, the release gate, Conventional Commits.
A skill is a directory with SKILL.md; the frontmatter needs name (equal to the directory), description and category. npm run validate enforces the contract, as CI does on every push.
npm test runs the validator self-test, the plugin-manifest check and the installer, store, doctor and cross-harness hook tests.
.claude-plugin/plugin.json and marketplace.json are generated by npm run plugin:build and checked by npm run plugin:check. They exist today; the Claude Code marketplace install path is still being finalised, so the installer above remains the supported route.
Releases are cut by release-please from Conventional Commits; do not bump package.json by hand. feat: means a minor bump.
Skills derived from other projects record source: in the frontmatter and an entry in THIRD_PARTY_NOTICES.md.

Found a bug or want a skill added? Open an issue.

aos

Set up

README

AOS — BDB Agent OS

Why AOS

What you get

Quick start

See it in action

Skills at a glance

What's new in v5

The dispatcher graph

Install

Profiles and options

Non-interactive

Local project harness

Supported harnesses

Capability tiers

The pipelines

The three variants side by side

Orchestrator chain

The GO gate

Gate profiles

Playbooks

Factory

How to run

A2A: sessions that talk to each other

Tools

AOS Hub and the sandbox

Fleet band and token-weather

MCP servers

MCP gateway and one MCP per app

Plugins and marketplace

AOS CLI

Memory and knowledge

What's included

The subagents (45)

Skills by category

Browse everything

Skills

Playbooks

Optional modules

memB: local vector memory

deja: transcript indexing

OpenWiki: living documentation

Rust + Tauri (rustauri)

StyleSeed (styleseed)

Codebase Memory MCP (codebase-memory)

Heimdall Token Saver: CLI output compression

Token optimization at a glance

Updating

Uninstall

FAQ

Contributing

Links

Rust + Tauri (`rustauri`)

StyleSeed (`styleseed`)

Codebase Memory MCP (`codebase-memory`)