@ranimontagna/agent-toolkit 0.1.28 → 0.1.30

package/CHANGELOG.md

@@ -1,5 +1,18 @@
 # Changelog
 
+## 0.1.30
+
+- onboard-repo: fixed the runtime-selection guidance — bare runtime flags
+  are mutually exclusive, so the skill now instructs
+  `--all-runtimes --no-gemini --no-antigravity` for multi-runtime installs.
+
+## 0.1.29
+
+- Added the bundled `onboard-repo` skill (`core` package): detects a
+  repository's stack, installs the matching toolkit skills, and organizes the
+  agent docs (canonical `AGENTS.md`, pointer `CLAUDE.md`, optional MCP
+  config), ending with a report of what was installed and skipped.
+
 ## 0.1.28
 
 - Frontend Skills: replaced Taste Skill with Vercel's Web Design Guidelines

package/README.md

@@ -178,6 +178,8 @@ skills/
   core/
     agent-toolkit-maintainer/
       SKILL.md
+    onboard-repo/
+      SKILL.md
   backend/
     api/
       api-design/
@@ -355,6 +357,7 @@ Follow these steps...
 | Package | Skill | Source |
 |---|---|---|
 | `core` | `agent-toolkit-maintainer` | Maintained in this repository |
+| `core` | `onboard-repo` | Maintained in this repository |
 | `backend` | `api-design` | Copied from Affaan Mustafa's [`ECC`](https://github.com/affaan-m/ECC/tree/main/skills/api-design) under the MIT license |
 | `backend` | `postgres-patterns` | Adapted from Affaan Mustafa's [`ECC`](https://github.com/affaan-m/ECC/tree/main/.kiro/skills/postgres-patterns) and [`sickn33/antigravity-awesome-skills`](https://github.com/sickn33/antigravity-awesome-skills/tree/main/skills/postgresql), both under the MIT license |
 | `backend` | `fastify-best-practices` | Copied from Matteo Collina's [`mcollina/skills`](https://github.com/mcollina/skills/tree/main/skills/fastify) under the MIT license |
@@ -599,6 +602,8 @@ skills/
   core/
     agent-toolkit-maintainer/
       SKILL.md
+    onboard-repo/
+      SKILL.md
   backend/
     fastify-best-practices/
       SKILL.md
@@ -688,8 +693,8 @@ Release a new npm version by updating `package.json`, pushing the change to
 `main`, then pushing a matching tag:
 
 ```bash
-git tag v0.1.28
-git push origin v0.1.28
+git tag v0.1.30
+git push origin v0.1.30
 ```
 
 For normal releases, prefer the scripted flow. It bumps `package.json`, updates

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ranimontagna/agent-toolkit",
-  "version": "0.1.28",
+  "version": "0.1.30",
   "description": "Personal AI agent toolkit installer for Claude Code, Codex CLI, OpenCode, Gemini CLI, Antigravity and graph-aware workflows.",
   "type": "module",
   "engines": {

package/skills/core/onboard-repo/SKILL.md

@@ -0,0 +1,89 @@
+---
+name: onboard-repo
+description: Use when setting up a repository for AI-assisted development - detect the stack, install the right skills through the Agent Toolkit, and organize the agent docs (AGENTS.md, CLAUDE.md, MCP config). Trigger phrases include "onboard this repo", "set up this project for AI", "which skills fit this repository", or "organize the agent config here".
+---
+
+# Onboard Repo
+
+Set up the current repository for AI-assisted development in three passes:
+detect the stack, install matching skills, and organize the agent docs.
+Report what you did and what you skipped at the end.
+
+Do not run installs or write files before completing the detection pass and
+showing the user your plan. Ask before overwriting any existing file.
+
+## Pass 1 — Detect the stack
+
+Inspect the repository (prefer fast reads over full scans):
+
+- Manifests: `package.json` (+ workspaces), `pyproject.toml`, `go.mod`,
+  `Cargo.toml`, `Gemfile`.
+- Frameworks and libraries in dependencies: React, React Native, Astro,
+  Next.js, Vue, GSAP, Tailwind, Playwright, Vitest/Jest.
+- Structure: monorepo (turbo/nx/pnpm workspaces), apps vs packages, CI
+  workflows, existing tests.
+- Existing agent config: `AGENTS.md`, `CLAUDE.md`, `.mcp.json`, `RTK.md`,
+  `.rtk/`, skill directories, `.agents/`, `.codex/` (never modify or delete
+  agent-state directories you did not create).
+
+Summarize findings in a few lines before proposing anything.
+
+## Pass 2 — Choose and install skills
+
+Map the detected stack to skills using this table, then install with a single
+Agent Toolkit run using only the flags that apply:
+
+| Signal | Skills | Toolkit selector |
+|---|---|---|
+| Any repo | grilling interviews, domain modeling | `--planning-skills-only` |
+| Frontend (React/Vue/Astro/CSS-heavy) | Impeccable, Web Design Guidelines, React Doctor | `--frontend-skills-only` |
+| Frontend | bundled `frontend` package (accessibility, design, gsap, react, react-native, astro — pick relevant) | `--skills-only` with `--skills-package frontend` |
+| Backend/API services | bundled `backend` package | `--skills-only` with `--skills-package backend` |
+| Infra/CI-heavy | bundled `devops` package | `--skills-only` with `--skills-package devops` |
+| Codebase feels tangled or user asks for audits | Improve | `--improve-only` |
+
+Notes:
+
+- Combine selections into as few toolkit invocations as possible; each
+  invocation is `npx -y @ranimontagna/agent-toolkit <flags>` plus runtime
+  selection. Bare runtime flags are mutually exclusive - to target several
+  runtimes use `--all-runtimes` with `--no-<runtime>` exclusions (default:
+  `--all-runtimes --no-gemini --no-antigravity`).
+- Skip anything already installed (check the runtime skill directories first).
+- Do not install skills for stacks the repo does not use; fewer, sharper
+  skills beat a pile of unused ones.
+
+## Pass 3 — Organize the agent docs
+
+Follow this layout (create lazily, ask before overwriting):
+
+- **`AGENTS.md`** — the canonical instruction file. Sections: project purpose
+  (one paragraph), stack and commands (dev/build/test/lint), repository rules
+  the agent must follow, available skills and when to reach for them, MCP
+  servers if any. Keep it under ~120 lines; link out instead of inlining long
+  docs.
+- **`CLAUDE.md`** — minimal pointer only:
+
+  ```markdown
+  # Claude Instructions
+
+  Read and follow `AGENTS.md` for this project.
+  ```
+
+- **`.mcp.json`** — only when the project benefits from specific MCP servers
+  (e.g. Playwright for browser validation, Firecrawl for research). Reference
+  env vars like `${FIRECRAWL_API_KEY}`; never hardcode secrets.
+- **`RTK.md` / `.rtk/`** — only when the user uses RTK in this repo.
+
+If an `AGENTS.md` already exists, update it (add the skills section, fix stale
+commands) instead of rewriting it. Preserve the existing voice and structure.
+
+## Final report
+
+End with a short report:
+
+1. Stack detected (one line).
+2. Skills installed (and for which runtimes) vs skipped-as-present.
+3. Files created or updated.
+4. Anything deliberately left out and why (e.g. "no MCP config — no browser
+   or research flows in this repo").