agent-marketplace

Garry Tan's gstack AI workflow skills, packaged as a Claude Code plugin with gstack: namespace. 47 skills including ship, review, qa, browse, investigate, office-hours, and more. Fork of garrytan/gstack.

59 skills

Joe Black's custom skills for Claude Code — agentic development, multi-provider plugin architecture, Claude Code plugin development, and best practices for working with Claude Code.

42 skills

Automatic mistake capture and proactive lesson injection for AI coding agents. Mines conversation logs for mistake patterns, structures them as indexed lessons, and injects relevant warnings before the agent repeats the mistake.

3 skills · 12 commands · 1 MCP server · hooks

Make every Claude Code session discoverable by what was done, asked, and questioned. Indexes transcripts into a sidecar SQLite store with heuristic and LLM-powered synthesis. Includes SessionEnd hook for forward capture and UserPromptSubmit hook for search reminders.

1 skill · 6 commands · hooks

Persistent context graph for codebase and web navigation. Symbol-chunks files, stores nodes in a SQLite graph, and returns compact references — enabling recall across sessions without re-reading files.

3 skills · 1 MCP server · hooks

Scaffold and maintain production-ready TypeScript projects — the cookiecutter-uv equivalent for the TypeScript ecosystem. Includes /new and /update commands plus a skill covering interactive and agent-driven scaffold flows.

1 skill · 2 commands · hooks

Scaffold and maintain production-ready Python projects — uv for dependency management, ruff + mypy for quality, pytest + tox-uv for testing, GitHub Actions CI, and Docusaurus docs. The cookiecutter for modern Python.

2 commands · hooks

Intercepts /compact and generates a structured session handoff instead of lossy compaction. Preserves critical decision context that would otherwise be lost during Claude Code's automatic context compaction.

1 command · hooks

Reference implementation of Anthropic's Code execution with MCP — reuses Claude Code's own sandbox to keep intermediate tool output out of the context window on every workflow. Up to 99% fewer tokens.

2 skills

Autonomous multi-agent PR reviewer — coordinates with other AI review bots to avoid duplication, runs 5+ specialized sub-agents, and self-improves through the full review → re-review → approve loop.

1 skill