{"id":"cdf5a115-8784-4700-a53c-49e69ef84f75","slug":"rldyour-opencode","kind":"mcp","name":"rldyour-opencode","summary":"rldyour AI CLI configuration for OpenCode: local plugins, MCP/LSP, permissions, commands, agents, browser/design workflows, and security review.","owner":"NDDev-Technocracy-Archive","repoUrl":"https://github.com/NDDev-Technocracy-Archive/rldyour-opencode","license":"AGPL-3.0","licenseTier":"strong-copyleft","trust":"community","stars":0,"match":null,"fewStars":true,"format":"mcp","createdAt":"2026-05-13T04:23:29.000Z","addedAt":"2026-10-21T06:05:54.260Z","logoUrl":"https://avatars.githubusercontent.com/u/340840341?v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["browser-web","design-media","memory-knowledge"],"tags":["agpl-3","ai-agents","ai-cli","browser-automation","commands","design-system","developer-tools","lsp"],"runtime":"binary","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-23T00:32:03.974Z","sample":false,"url":"https://conduits.dev/tools/rldyour-opencode","language":"Python","transports":["stdio"],"install":{},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# rldyour-opencode\n\n`rldyour-opencode` is the rldyour AI CLI configuration for OpenCode: local plugins, MCP/LSP, permissions, commands, agents, browser/design workflows, and security review.\n\n[![validate](https://github.com/rldyourmnd/rldyour-opencode/actions/workflows/validate.yml/badge.svg?branch=main)](https://github.com/rldyourmnd/rldyour-opencode/actions/workflows/validate.yml)\n[![CodeQL](https://github.com/rldyourmnd/rldyour-opencode/actions/workflows/codeql.yml/badge.svg?branch=main)](https://github.com/rldyourmnd/rldyour-opencode/actions/workflows/codeql.yml)\n[![OpenSSF Scorecard](https://api.scorecard.dev/projects/github.com/rldyourmnd/rldyour-opencode/badge)](https://scorecard.dev/viewer/?uri=github.com/rldyourmnd/rldyour-opencode)\n[![License: AGPL-3.0-or-later](https://img.shields.io/badge/License-AGPL--3.0--or--later-blue.svg)](https://www.gnu.org/licenses/agpl-3.0)\n[![Latest Release](https://img.shields.io/github/v/release/rldyourmnd/rldyour-opencode)](https://github.com/rldyourmnd/rldyour-opencode/releases/latest)\n\n`rldyour-opencode` is the rldyour AI CLI configuration for OpenCode: local TypeScript plugins, MCP/LSP, permissions, slash commands, subagents, browser/design workflows, and security review. Russian-first SDLC workflow, Serena integration, full-auto owner posture - all native to the OpenCode AI coding agent format.\n\n## Current Baseline\n\n| Field | Value |\n|---|---|\n| Adapter version | `1.7.33` |\n| Runtime baseline | OpenCode 1.17.18 |\n| GitHub release tag | `1.7.33` |\n\nRuntime baseline source: `references/opencode-baseline.json`. Submodule pins are owned by the root control-plane `config/repositories.json`.\n\nValidated against `opencode-ai`, `@opencode-ai/plugin`, and\n`@opencode-ai/sdk` `1.17.18`. The v1.14.48 → v1.17.18 jump preserves the\nruntime hook surface and tool-ID naming while picking up current plugin\nloading, skill discovery, ACP/WebSocket reliability, safer edits,\nbackgroundable subagents, Copilot zero-batch pricing protection, and the Meta\nMuse Spark system prompt. The `1.17.18` config schema is byte-identical to the\n`v1.17.7`/`v1.17.13`/`v1.17.14` snapshots (SHA-256 `57c02429`).\n\n## What This Repository Provides\n\n`rldyour-opencode` is a self-contained OpenCode project configuration package - not a fork of the OpenCode runtime. Drop `opencode.json` and `.opencode/` into any project and OpenCode resolves the full configuration: 38 skills across 10 workflow domains, 9 subagents, 11 slash commands, 10 TypeScript plugins, 11 MCP servers, 8 custom LSP servers, and owner-standard full-auto permissions. The adapter is authored by Danil Silantyev (github:rldyourmnd), CEO NDDev, and is licensed under AGPL-3.0-or-later. Implementation changes belong in this repository; the control-plane superproject (`rldyour-ai-cli-tools`) only advances the submodule pin.\n\n## Native Boundaries\n\nOpenCode's native config surfaces that this adapter populates:\n\n- **Master config**: `opencode.json` (JSON or JSONC) - model, MCP, LSP, agents, watchers, compaction, permission rules.\n- **Release-safe overlay**: `opencode.release-safe.json` - conservative static read-deny patterns for `.env`, private keys, tokens, and shell/edit ask posture for public OSS examples.\n- **Durable AI context**: `AGENTS.md`, `.opencode/`, `.serena/project.yml`, and `.serena/memories/` are tracked on `main`. Runtime-local Serena state remains ignored.\n- **`.opencode/` directory layout**:\n  - `agents/*.md` - 9 subagents (6 reviewer tracks, memory-sync, ry-explore, customize-opencode)\n  - `skills/<name>/SKILL.md` - 38 skills across 10 domains\n  - `commands/*.md` - 11 slash commands\n  - `plugins/*.ts` - 10 Bun-runtime TypeScript local plugins\n  - `package.json` - `@opencode-ai/plugin` pin for local Bun dependency resolution\n- **Permission keys (canonical v1.17.18)**: `read`, `edit`, `bash`, `task`, `external_directory`, `doom_loop` - used in `opencode.json` `permission.*` fields.\n- **MCP JSON**: declared under `mcp` in `opencode.json`; registry-backed local\n  servers use `bunx`/`uvx`, Dart uses its SDK, and Chrome DevTools uses the\n  bootstrap-managed CloakBrowser wrapper - never `npx`.\n\nSource-only artifacts (scripts, tests, CI workflows, reference docs, ADRs) are not loaded by OpenCode at runtime; they exist for validation and release hygiene only.\n\n## Install / Update / ry-repair\n\n**Install** - clone and copy into your project:\n\n```bash\ngit clone https://github.com/rldyourmnd/rldyour-opencode.git\ncd rldyour-opencode\ncp opencode.json /path/to/your/project/opencode.json\ncp -r .opencode /path/to/your/project/.opencode\ncp AGENTS.md /path/to/your/project/AGENTS.md\n```\n\n**Authenticate** - log in to your provider interactively, then set any required MCP env vars:\n\n```bash\nopencode auth login\nexport GITHUB_PERSONAL_ACCESS_TOKEN=YOUR_PLACEHOLDER_TOKEN  # required for GitHub MCP\nexport CONTEXT7_API_KEY=YOUR_PLACEHOLDER_KEY               # optional\n```\n\n**Run**:\n\n```bash\ncd /path/to/your/project\nopencode\n```\n\n**Check resolved config** (authoritative):\n\n```bash\nopencode debug config\n```\n\n**Diagnose runtime** (MCP, LSP binaries, agent/skill/command discovery, git):\n\n```bash\nbash scripts/doctor_opencode.sh\n```\n\n**Owner launcher** (`oc`) - the `rldyour-ai-cli-tools` root provides `scripts/install_yolo_launchers.sh --apply`, which installs the `oc` wrapper that injects an allow-all `OPENCODE_CONFIG_CONTENT` override and sets `OPENCODE_DISABLE_CLAUDE_CODE=1` so root skill resolution uses `.opencode/skills`. Use `oc` for the trusted owner workstation full-auto posture.\n\n**Convergence** - inside any project where this config is active, run `/ry-repair` to repair stale docs, memories, contracts, hooks, MCP/LSP config, CI, and AI-tool context.\n\n**Update** - pull the latest adapter tag, re-copy `opencode.json` and `.opencode/`, then run `opencode debug config` to confirm runtime resolved correctly. Check `CHANGELOG.md` and `references/opencode-baseline.json` for any dependency bumps that require re-authentication or pin updates.\n\n## Active Catalog\n\n| Layer | Where | Count |\n|---|---|---|\n| Master config | `opencode.json` | 1 |\n| Release-safe overlay | `opencode.release-safe.json` | 1 |\n| Cross-tool instructions | `AGENTS.md` | 1 |\n| Claude Code project memory (agent-only) | `.claude/CLAUDE.md` | 1 |\n| Subagents | `.opencode/agents/*.md` | 9 |\n| Skills | `.opencode/skills/<name>/SKILL.md` | 38 |\n| Slash commands | `.opencode/commands/*.md` | 11 |\n| Plugins | `.opencode/plugins/*.ts` | 10 |\n| Custom diagnostic tools (LLM-callable) | `.opencode/plugins/ry-tools.ts` | 5 |\n| MCP servers | `opencode.json` → `mcp` | 11 |\n| Custom LSP servers | `opencode.json` → `lsp` | 8 |\n| Reference docs (contracts + machine metadata) | `references/*` | 22 |\n| Operator guides | `docs/*.md` | 5 |\n| Architecture decision archive | `docs/decisions/*.md` | 10 |\n| Diagnostic scripts (bash + Python) | `scripts/` | 30 |\n| Pytest suites | `scripts/tests/*.py` | 26 |\n| CI workflows | `.github/workflows/*.yml` | 11 |\n\n### Slash Commands\n\n| Command | Agent | Purpose |\n|---|---|---|\n| `/ry-init` | `build` | Scoped read-only project context with Serena-first discovery |\n| `/ry-start` | `build` | Full task lifecycle: init → research → plan → implement → verify → sync |\n| `/ry-review` | `plan` | Report-only deep review with parallel reviewer subagents |\n| `/ry-repair` | `build` | Repair stale docs, memories, contracts, hooks, MCP/LSP config, CI, and AI-tool context |\n| `/ry-newp` | `build` | Plan a new project (skeptical questions, research, ADRs, architecture docs) |\n| `/ry-deploy` | `build` | Deploy with sync, log checks, fix-forward |\n| `/ry-sync` | `build` | Synchronize memories, docs, git, and tracked context |\n| `/ry-design` | `build` | End-to-end design: Figma → tokens → FSD → shadcn/ui → browser validation |\n| `/ry-explore` | `ry-explore` (subtask) | Deep multi-source research via Context7 / DeepWiki / Grep / web |\n| `/ry-sec-review` | `plan` | Defensive Mythos-style security review |\n| `/ry-rules-review` | `plan` | Audit implementation against rldyour rules (report-only) |\n\n### MCP Servers\n\nLocal servers timeout 30 s, remote 15 s. Launcher convention: `bunx` for npm,\n`uvx` for Python, `dart` for Dart SDK, and the exact managed wrapper for Chrome\nDevTools.\n\n| Server | Type | Version | Purpose |\n|---|---|---|---|\n| serena | local (uvx) | 1.5.3 | Semantic code navigation, analysis, editing |\n| sequential-thinking | local (bunx) | 2026.7.4 | Structured reasoning |\n| chrome-devtools | local (managed CloakBrowser wrapper) | 1.5.0 | Chrome DevTools diagnostics |\n| shadcn | local (bunx) | 4.13.0 | shadcn/ui registry access |\n| dart-flutter | local (dart) | - | Dart/Flutter project support |\n| context7 | remote | - | Current library documentation |\n| deepwiki | remote | - | Repository documentation |\n| grep | remote | - | Search across public GitHub repos |\n| figma | remote | - | Figma design context |\n| github | remote | toolsets: context,repos,issues,pull_requests,users | Remote GitHub MCP endpoint (requires PAT) |\n| openai-docs | remote | - | Official OpenAI/Codex documentation |\n\n### TypeScript Plugins\n\n10 Bun-runtime plugins loaded from `.opencode/plugins/*.ts`:\n\n- **Lifecycle**: `ry-bootstrap` (session banner + compaction context + autocontinue), `ry-env-protection` (block sensitive reads with toast), `ry-shell-strategy` (shell env + git push guardrails), `ry-sync-reminder` (idle toast), `ry-flow-hooks` (commit advice + post-commit nudge).\n- **LLM-side**: `ry-tools` (5 custom diagnostic tools the LLM can call), `ry-command-audit` (credential-sanitized slash-command audit log), `ry-tool-hints` (routing nudges injected into MCP tool descriptions).\n- **Runtime context + permission events**: `ry-system-context` (date + branch + HEAD SHA + dirty state injected into every system prompt), `ry-permission-events` (observability-only `permission.asked` / `permission.replied` event audit).\n\n### LSP Servers\n\n8 custom LSP servers on top of OpenCode's 35+ built-ins: `ruff`, `vscode-html`, `vscode-css`, `vscode-json`, `docker`, `taplo`, `marksman`, `qmlls`.\n\n### Models\n\nThe marketplace ships with `moonshotai/kimi-k2.7-code` as the top-level default. Subagents inherit the top-level model.\n\n| Slot | Default in this repo | Common Anthropic alternative |\n|---|---|---|\n| `model` (primary) | `moonshotai/kimi-k2.7-code` | `anthropic/claude-sonnet-4-6` |\n| `small_model` | `moonshotai/kimi-k2.7-code` | `anthropic/claude-haiku-4-5-20251001` |\n| `default_agent` | `build` | `build` |\n| Reviewer / memory-sync / explore subagents | inherit top-level `model` | inherit top-level `model` |\n\nRun `opencode models <provider>` to list every accepted ID. To switch provider, edit `\"model\"` in `opencode.json` and confirm with `opencode debug config`.\n\n## Browser / Design / DevTools Routing\n\nTwo browser execution providers are active. Every browser action first runs\nexact `$HOME/.local/bin/cloakbrowser-cdp-health`; missing or nonzero health\nstops with `NOT_PROVEN` and no fallback:\n\n- **Managed Playwright CLI** (`$HOME/.local/bin/playwright-cli`) - UI and flow\n  evidence. `run-code`, `--filename`, raw Playwright, package runners, and\n  alternate executables/configs are forbidden.\n- **Managed Chrome DevTools MCP** (`chrome-devtools`, version 1.5.0) - console,\n  network, runtime, layout, performance, Lighthouse, and memory diagnosis\n  through the exact CloakBrowser wrapper in `opencode.json`.\n\n`webwright-task` remains as a compatibility skill name for long-horizon intent,\nbut routes to these providers and never executes the Webwright Python runtime.\nStock/raw/in-app Browser, `browser_agent`, `node_repl`, computer-use, Playwright\nMCP, alternate CDP endpoints, direct packages, and all fallbacks are forbidden.\n\nThe `/ry-design` command routes through Figma MCP (design context and asset download), shadcn/ui MCP (registry access), and Chrome DevTools MCP (validation). DeepWiki, Context7, and Grep MCP support research and documentation retrieval during design and exploration tasks.\n\n## Repository Context / Serena Memory\n\nNormal `main` history carries product artifacts and durable agent context:\n`opencode.json`, `.opencode/`, `AGENTS.md`, `.serena/project.yml`,\n`.serena/memories/*.md`, scripts, tests, CI workflows, docs, and reference\nfiles.\n\nSerena memories live under `.serena/memories/` using the `AREA-NN-SLUG.md`\ntaxonomy. The freshness contract: memories are updated only from verified\ncurrent code, git diffs, and tests - never from speculation, plans, or chat\nhistory. Runtime-local cache, reviews, diagnostics, markers, local env files,\nbrowser artifacts, tokens, cookies, and credentials stay ignored.\n\n## Security Boundary\n\nOwner full-auto posture is intentional and explicitly acknowledged. The primary `build` and `plan` agents use OpenCode's canonical v1.17.18 permission keys with `\"allow\"` for `read`, `edit`, `bash`, `task`, `external_directory`, and `doom_loop`. This is not a sandbox - it is a trusted owner workstation posture designed for maximum autonomy.\n\nThe `oc` launcher (from `scripts/install_yolo_launchers.sh --apply`) injects an allow-all `OPENCODE_CONFIG_CONTENT` environment override and sets `OPENCODE_DISABLE_CLAUDE_CODE=1`, mirroring the same no-prompt posture at the OS launcher level.\n\nReviewer subagents are explicitly stricter: `edit: \"deny\"`, `bash` allowlists limited to read-only git verbs. Their role contract is report-only review, not implementation.\n\nThe `ry-env-protection` plugin blocks reads of `.env`, private key files, tokens, and credentials with a toast notification before the LLM sees them. The `opencode.release-safe.json` overlay provides a conservative alternative profile for public OSS examples and cautious installs.\n\nMCP secrets (`GITHUB_PERSONAL_ACCESS_TOKEN`, `CONTEXT7_API_KEY`, etc.) are passed as environment variables and are never committed. Use `.env.example` as a reference; never populate `.env` with real values in version control. Security vulnerabilities should be reported via GitHub Security Advisories - do not file public issues.\n\n## Validation\n\n```bash\nbash scripts/validate_config.sh                            # JSON shape + skill/agent/command frontmatter + VERSION semver\nuvx --from \"pytest==9.1.1\" --with \"pyyaml==6.0.3\" --with \"jsonschema==4.26.0\" --with \"referencing==0.37.0\" pytest scripts/tests/\nbash scripts/check_deps_freshness.sh --check-freshness     # list pinned MCP dependencies + npm/PyPI freshness\npython3 scripts/check_action_pins.py .github/workflows --remote  # verify SHA-pinned GitHub Actions comments\npython3 scripts/check_plugin_hooks.py                      # verify plugin hook contract; forbids permission.ask as enforcement\npython3 scripts/validate_contract.py                       # verify canonical rldyour adapter contract\npython3 scripts/smoke_mcp_capabilities.py                  # probe every MCP server for reachability\npython3 scripts/validate_instruction_docs.py               # verify AGENTS.md + .claude/CLAUDE.md anchor headings\nbash scripts/doctor_opencode.sh                            # full diagnostics: MCP, LSP binaries, agent/skill/command discovery, git\nbash scripts/check_lsps.sh                                 # 16 language servers + project prereqs\nbash scripts/collect_diagnostics.sh --include-doctor       # local timestamped diagnostic bundle for triage\nopencode debug config                                      # native resolved config (authoritative)\nopencode debug agent <name>                                # validate individual agent\nopencode models anthropic                                  # list available models for the active provider\n```\n\nCI mirrors core checks via `.github/workflows/validate.yml` on every push/PR to `main`. `.github/workflows/dependency-check.yml` runs weekly to surface MCP pin freshness. Public repositories use standard GitHub-hosted runners; keep `share: \"manual\"` unchanged (it controls OpenCode session sharing, not CI execution). See `docs/observability.md` for the full triage flow.\n\nCommands marked NOT_PROVEN (e.g., live MCP probes) require network access and installed binaries; skip them in offline or CI-only environments.\n\n## Release / Rollback\n\nReleases are tag-driven. Every product version must have a matching numeric GitHub Release at `github.com/rldyourmnd/rldyour-opencode/releases`. A `VERSION` file update alone is not sufficient.\n\nVersion movement follows the control-plane policy: default is patch (`+0.0.1`) after a public release exists; minor and major bumps are owner-directed decisions only. `CHANGELOG.md` follows Keep a Changelog 1.1.0. `SECURITY.md` lists the exact supported version tag.\n\n**Rollback**: to revert to a prior release, check out the numeric tag (`git checkout X.Y.Z`), re-copy `opencode.json` and `.opencode/` into your project, and re-run `opencode debug config`. See `docs/rollback-restore.md` for the full rollback and restore procedure. Dependency update policy is in `docs/dependency-updates.md`.\n\n## Support / License\n\n- **License**: AGPL-3.0-or-later\n- **Author**: Danil Silantyev (github:rldyourmnd), CEO NDDev\n- **Issues**: [github.com/rldyourmnd/rldyour-opencode/issues](https://github.com/rldyourmnd/rldyour-opencode/issues) - bug reports, regression evidence, missing-component requests.\n- **Discussions**: [github.com/rldyourmnd/rldyour-opencode/discussions](https://github.com/rldyourmnd/rldyour-opencode/discussions) - general questions, workflow advice, configuration clarifications, sharing usage patterns.\n- **Releases**: [github.com/rldyourmnd/rldyour-opencode/releases](https://github.com/rldyourmnd/rldyour-opencode/releases) - numeric product tags (`X.Y.Z`) with release notes from CHANGELOG.\n- **Security**: Report vulnerabilities via [GitHub Security Advisories](https://github.com/rldyourmnd/rldyour-opencode/security/advisories) - do not file public issues for security vulnerabilities.\n\nThis is a personal marketplace - response time is best-effort, no SLA. Feel free to fork and tailor to your own workflow.\n"}