ai-agent-skills

╔═══════════════════════════════════════════════════════════════╗
║                                                               ║
║  █████╗ ██╗     █████╗  ██████╗ ███████╗███╗   ██╗████████╗   ║
║  ██╔══██╗██║    ██╔══██╗██╔════╝ ██╔════╝████╗  ██║╚══██╔══╝  ║
║  ███████║██║    ███████║██║  ███╗█████╗  ██╔██╗ ██║   ██║     ║
║  ██╔══██║██║    ██╔══██║██║   ██║██╔══╝  ██║╚██╗██║   ██║     ║
║  ██║  ██║██║    ██║  ██║╚██████╔╝███████╗██║ ╚████║   ██║     ║
║  ╚═╝  ╚═╝╚═╝    ╚═╝  ╚═╝ ╚═════╝ ╚══════╝╚═╝  ╚═══╝   ╚═╝     ║
║  ███████╗██╗  ██╗██╗██╗     ██╗     ███████╗                  ║
║  ██╔════╝██║ ██╔╝██║██║     ██║     ██╔════╝                  ║
║  ███████╗█████╔╝ ██║██║     ██║     ███████╗                  ║
║  ╚════██║██╔═██╗ ██║██║     ██║     ╚════██║                  ║
║  ███████║██║  ██╗██║███████╗███████╗███████║                  ║
║  ╚══════╝╚═╝  ╚═╝╚═╝╚══════╝╚══════╝╚══════╝                  ║
║                                                               ║
║       Public Agent Skills collection for coding agents        ║
║    Cursor - Claude Code - Codex - Windsurf - and more ...     ║
║    goal-mode - memo-session-skill - memo-session-mcp - MIT    ║
║    ai-software-factory-pipeline · github.com/shenwell/...     ║
║            npx skills add shenwell/ai-agent-skills            ║
║                                                               ║
╚═══════════════════════════════════════════════════════════════╝

Public Agent Skills for Cursor, Claude Code, Codex, Windsurf, and more.

• goal-mode — autonomous coding agent, verify until done: verifiable finish line, worker⇄verifier proof loop (open alternative to Claude Code /goal for Cursor).
• memo-session-skill — persistent AI agent memory for coding sessions: cross-session knowledge survives context resets via git-tracked memory/, wiki, and handoffs; pairs with goal-mode checkpoints.
• memo-session-mcp — MCP search companion for memo-session: FTS index over portfolio global-memory, project memories, and optional PDF/Excel corpora — use from any Cursor chat.
• AI Software Factory Pipeline — multi-agent software pipeline for Cursor. Classifier → analyst → implementer → reviewer moves chat or GitHub issues to a feature branch and PR. Separate repo, not installed via npx skills add.

Stops the common failure mode: the agent tries once, claims “done,” and leaves a red build.

shenwell/ai-agent-skills
└── skills/
    ├── goal-mode/              ← keep going until tests/lint/build are green
    ├── memo-session-skill/     ← persistent agent memory · session → memory/wiki/handoff
    ├── memo-session-mcp/       ← MCP FTS search · portfolio + project memory + docs
    └── <next-skill>/

Related (separate repo): [ai-software-factory-pipeline](https://github.com/shenwell/ai-software-factory-pipeline) — `/factory` multi-agent pipeline for Cursor.

Quickstart

1. Install

Global — available in all projects on this machine:

npx skills add shenwell/ai-agent-skills --skill goal-mode -g -a cursor -y

This repository only — install into the current project (share with the team via git):

npx skills add shenwell/ai-agent-skills --skill goal-mode -a cursor -y

-a cursor installs for Cursor; -y skips the interactive agent picker.
Other agents: add more -a flags, e.g. -a cursor -a claude-code -a codex.
List skills: npx skills add shenwell/ai-agent-skills --list

Recommended pair for long autonomous runs:

npx skills add shenwell/ai-agent-skills --skill goal-mode -g -a cursor -y
npx skills add shenwell/ai-agent-skills --skill memo-session-skill -g -a cursor -y

Optional — memo-session-mcp (search portfolio + project memory from any chat):

cd skills/memo-session-mcp && pip install -e .
memo-session-index reindex   # after setting GLOBAL_MEMORY_ROOT

Register in ~/.cursor/mcp.json — see skills/memo-session-mcp/README.md.

2. First run (in any project)

Open the project in Cursor (or your agent) and run:

/goal Fix all ESLint errors in src; tests and build must pass

On the first /goal in a project the agent bootstraps automatically: copies goal.config.yml, hooks, /goal command, templates into .cursor/. You do not need a separate bootstrap command for normal use.

If you want hooks/config before the first /goal, or the agent could not find the skill path:

# pick the path that exists after install
node ~/.cursor/skills/goal-mode/scripts/goal-bootstrap.js
node ~/.agents/skills/goal-mode/scripts/goal-bootstrap.js
# Windows PowerShell:
node $env:USERPROFILE\.cursor\skills\goal-mode\scripts\goal-bootstrap.js

Who this is for

• Engineers who want an agent to keep going until green (lint, tests, CI), not stop after one attempt
• Teams using Cursor Cloud Agent or long unattended runs
• Anyone looking for an open Claude Code /goal alternative with a durable contract + verifier

Available skills

goal-mode — autonomous coding agent, verify until done

Keep the agent working until a verifiable finish line — tests green, lint clean, build passing, migration done — with a GOAL.md contract, worker⇄verifier proof loop, and auto-resume. Open alternative to Claude Code /goal for Cursor and other hosts.

Package: skills/goal-mode/ · skill README: skills/goal-mode/README.md

  YOU                          GOAL MODE                         DONE
   │                               │                               │
   │  /goal Fix lint errors        │                               │
   ├──────────────────────────────►│  intake → plan → execute      │
   │                               │  worker ⇄ verifier            │
   │                               ├──────────────────────────────►│
   │                               │         COMPLETE + report     │

┌─────────────────────────────────────────────────────────────────────────┐
│  CONTRACT   goals/{id}/GOAL.md + goal.config.yml                        │
│  AGENTS     intake → planner → phase-planner → worker → verifier        │
│  HOOKS      sessionStart + stop (auto-continue)                         │
│  SCRIPTS    init · status · verify · time · bootstrap                   │
└─────────────────────────────────────────────────────────────────────────┘

Configure after first /goal or manual bootstrap (.cursor/goal.config.yml):

verify:
  test: { command: "npm test", required: true }
  lint: { command: "npm run lint", required: true }
budget:
  max_iterations: 50
  max_hours: 8
memory:
  skill: memo-session-skill
  checkpoints:
    enabled: true
    on_phase_complete: full
    on_complete: full

memo-session-skill — persistent AI agent memory

Cross-session persistence for coding agents: consolidate session knowledge and route it into MEMORY.md, memory/ (HOT/WARM/COLD), project wiki, AGENTS.md, skills, and optional portfolio memory — git-tracked write-path memory with conflict gate and temperature limits. Complements RAG and managed platforms (Mem0, Zep); does not replace them.

Package: skills/memo-session-skill/ · skill README: skills/memo-session-skill/README.md

  SESSION                    MEMO SESSION                      NEXT TURN
     │                            │                                │
     │  wrap up the session       │  preflight → classify → route  │
     ├───────────────────────────►│  HOT / WARM / wiki / portfolio │
     │                            ├───────────────────────────────►│
     │                            │         handoff + changelog    │

Install:

npx skills add shenwell/ai-agent-skills --skill memo-session-skill -g -a cursor -y

Integration with goal-mode: memo-session goal-mode-integration

AI Software Factory Pipeline

Four stations (classifier, analyst, implementer, reviewer) plus optional researcher and designer. Work enters from chat or a labeled GitHub issue; the pipeline opens a feature branch and pull request. Merge stays with you.

Repository: github.com/shenwell/ai-software-factory-pipeline

This is a workspace kit in its own repo, not a single npx skills add drop-in from this collection.

git clone https://github.com/shenwell/ai-software-factory-pipeline.git

Open the clone in Cursor, run /factory-setup, then /factory <task> or /factory #N.

Add another skill

skills/my-skill/SKILL.md   # name + description frontmatter (required)

1. Create the folder + SKILL.md
2. npx skills add . --list
3. Document it above
4. Push → npx skills add shenwell/ai-agent-skills --skill my-skill

See CONTRIBUTING.md.

Layout

README.md
LICENSE
CONTRIBUTING.md
.gitignore
.github/ISSUE_TEMPLATE/
skills/
├── goal-mode/
├── memo-session-skill/
└── memo-session-mcp/

Related: ai-software-factory-pipeline (separate repository).

License

MIT — LICENSE. Not affiliated with Anthropic or Claude Code.