A curated collection of skills, extensions, and instructions for AI coding agents. Skills sync to a shared ~/.agents/skills directory and are symlinked into Claude Code, Pi, Codex, and other runtimes; _AGENTS.md is the single instruction file linked to all of them.
Requires mise.
git clone --recursive https://github.com/vaayne/agent-kit.git
cd agent-kit
mise run setup # submodules, dependencies
mise run sync # everything: skills + instructions + extensions
mise run sync:skills # local + remote skills → ~/.agents/skills, symlinked into each runtime
mise run sync:agents # _AGENTS.md → CLAUDE.md / AGENTS.md symlinks for every framework
mise run sync:extensions # Pi extensions → ~/.pi/agent/extensions symlinksThe core of the kit is a development lifecycle where every step is an independently invocable skill, and spec-dev is the thin orchestrator over them:
scout → grill → blueprint → mason → code-review → teach
探地形 拷问 画图纸 施工 审计 内化
| Skill | Role |
|---|---|
| scout | Find your unknowns before they get expensive — blindspot pass, prototypes, references, quadrant diagnostic |
| grill | Stress-test an idea through structured interrogation, one question at a time |
| blueprint | Write a decision-first plan.md — decisions with tradeoffs, phased tasks with acceptance blocks |
| mason | Execute a plan.md phase by phase: catch up → implement → verify → commit → handoff |
| code-review | Multi-perspective adversarial review with verifier subagents and near-zero false positives |
| teach | Socratic quiz loop — merge only what you can pass a quiz on |
| spec-dev | The orchestrator: sequence, review gates, and skip rules over the skills above |
| delegate | Run a self-contained task in a separate agent session (Claude Code or Pi, auto-routed by model name) |
| refine-code | Improve existing code without changing behavior — simplify, deepen abstractions, reduce complexity |
| handoff | Transfer context to a fresh focused session |
| Skill | Description |
|---|---|
| curator | Maintain the nmem knowledge base — lint + synthesis passes |
| tap-web | Web access, search, extraction, and browser automation via tap |
| designer | Distinctive UI/UX design and prototypes for filesystem-backed agents |
| humanizer | Strip AI writing patterns from prose |
| kreuzberg | Extract text/tables/metadata from 91+ document formats |
| python-script | Robust Python automation with logging and safety checks |
| vertex-ai-image | Image generation, editing, and understanding via Google Gemini |
| cf-email | Send email through the Cloudflare Email Sending API |
| gws | Google Workspace operations via the gws CLI |
| lark | Lark/Feishu workspace operations via lark-cli |
| openlist | Manage files on OpenList/AList cloud storage |
Remote skills installed during sync (see skills/remote-skills.txt): skill-creator, mcp-skill-gen, native-feel-skill.
| Extension | Description |
|---|---|
| delegate | Delegate tasks to specialized presets |
| cliproxy-provider | Register a cliproxy provider from a /v1/models endpoint |
| firework-provider | Fireworks provider with router model and tool-field sanitizing |
| codex-usage-status | Show Codex usage windows in the status line |
| model-context | Tell the agent its active Pi model without changing the prompt |
| notify | Play a sound when the agent finishes a task |
Delegate presets: librarian (code research), oracle (architecture analysis), reviewer (correctness/risk review), search (fast retrieval), ui-engineer (visual/UI), worker (general execution).
agent-kit/
├── _AGENTS.md # Shared agent instructions, symlinked to every framework
├── skills/ # Local skills + remote-skills.txt registry
├── extensions/ # Pi extensions
└── mcphub # MCP hub (submodule)