Skip to content
Back to search

my-ai-tools

MarketplaceWorks with Claude CodeGitHub

Marketplace for my-ai-tools plugins

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/jellydn/my-ai-tools with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/jellydn/my-ai-tools'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/jellydn--my-ai-tools/badge.svg)](/listings/jellydn--my-ai-tools)

HTML

<a href="/listings/jellydn--my-ai-tools"><img src="/api/listings/jellydn--my-ai-tools/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

These steps run a script from the network that findsafeskills does not scan.

The installer can replace existing configuration files. Run a dry run first and use `--backup` or `-Backup` when you want a restorable copy.

macOS, Linux, and WSL

Review the installer, then run it:

curl -fsSL https://ai-tools.itman.fyi/install.sh -o install.sh
less install.sh
bash install.sh --dry-run
bash install.sh --backup

For a direct one-line installation:

curl -fsSL https://ai-tools.itman.fyi/install.sh | bash

Windows 11

Open PowerShell and run the canonical installer:

Invoke-RestMethod https://ai-tools.itman.fyi/install.ps1 | Invoke-Expression

This downloads and executes a remote script. Review [install.ps1](install.ps1) before trusting it. It installs missing prerequisites through `winget`, reuses existing tools, refreshes PATH for the current process, and starts the shared installer in Git Bash. Windows may request administrator approval for Git or Node.js installation. Package installation uses `--source winget`, not the Microsoft Store. Node and npm are discovered again after installation and passed to Git Bash in the same process; no separate prerequisite command or terminal restart is needed after a successful install. The wrapper skips Git Bash login profiles so an older shell PATH cannot hide these tools. Approve only the package installer's elevation prompt; the wrapper does not relaunch the full configuration installer as administrator. If approval is denied, setup stops and can be rerun safely. Native Windows uses Node/npm, not Bun, for package installation; Linux, macOS, and WSL keep Bun-first behavior.

The installer can replace existing configuration. To inspect the script, preview changes, and request a backup, use the saved-script path instead:

Invoke-RestMethod https://ai-tools.itman.fyi/install.ps1 -OutFile install.ps1
Get-Content .\install.ps1
powershell -NoProfile -ExecutionPolicy Bypass -File .\install.ps1 -DryRun
powershell -NoProfile -ExecutionPolicy Bypass -File .\install.ps1 -Backup

Execution-policy bypass applies only to that process. `-DryRun` does not install prerequisites; install them first if the preview reports they are missing. The one-line command is non-interactive and automatically uses `--yes`.

If setup fails, resolve the reported error and rerun the same command. If a newly installed prerequisite still cannot be found, close all terminal windows and reopen PowerShell. If `winget` is missing, install or update [App Installer](https://aka.ms/getwinget), or install [Git for Windows](https://git-scm.com/download/win), [`jq`](https://jqlang.org/download/), and [Node.js LTS with npm](https://nodejs.org/) manually. In Git setup, select **Git from the command line and also from 3rd-party software**. Do not bypass package lifecycle errors with `--ignore-scripts`: native binaries can remain uninstalled.

If the wrapper says all prerequisites are met but Bash then reports missing Node/npm, check that the downloaded `install.ps1` includes `Find-Node` and the `OpenJS.NodeJS.LTS` installation step. A stale web deployment can serve an older wrapper while cloning the latest Bash installer. Maintainers must check the **Dokku Deploy** workflow for `ai-tools.itman.fyi`; a successful **Deploy GitHub Pages** run does not update that domain. Until the domain is updated, use a reviewed repository checkout and run:

powershell -NoProfile -ExecutionPolicy Bypass -File .\install.ps1 -Yes

Repository checkout

git clone https://github.com/jellydn/my-ai-tools.git
cd my-ai-tools
./cli.sh --dry-run
./cli.sh --backup

On Windows PowerShell, use the native wrapper from the checkout:

git clone https://github.com/jellydn/my-ai-tools.git
Set-Location my-ai-tools
.\install.ps1 -DryRun
.\install.ps1 -Backup

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 267 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. {
  2. "$schema": "https://anthropic.com/claude-code/marketplace.schema.json",
  3. "name": "my-ai-tools",
  4. "description": "Marketplace for my-ai-tools plugins",
  5. "owner": {
  6. "name": "my-ai-tools",
  7. "email": "dung@productsway.com"
  8. },
  9. "plugins": [
  10. {
  11. "name": "accountable-engineering",
  12. "description": "Guides disciplined AI-assisted engineering that avoids cognitive surrender and keeps humans accountable. Use for non-trivial implementation, architecture, security, or operational tasks.",
  13. "version": "1.0.0",
  14. "author": {
  15. "name": "Dung Huynh Duc",
  16. "email": "dung@productsway.com"
  17. },
  18. "source": "./skills/accountable-engineering",
  19. "category": "productivity"
  20. },
  21. {
  22. "name": "babysit-pr",
  23. "description": "Monitors an open GitHub PR for CI failures, review feedback, mergeability, and safe retries or fixes.",
  24. "version": "1.0.0",
  25. "author": {
  26. "name": "my-ai-tools contributors"
  27. },
  28. "source": "./skills/babysit-pr",
  29. "category": "productivity"
  30. },
  31. {
  32. "name": "adr",
  33. "description": "Manages Architecture Decision Records (ADR) for tracking important architectural decisions. Initialize, create, list, search, and view ADRs.",
  34. "version": "1.0.0",
  35. "author": {
  36. "name": "Dung Huynh Duc",
  37. "email": "dung@productsway.com"
  38. },
  39. "source": "./skills/adr",
  40. "category": "productivity"
  41. },
  42. {
  43. "name": "codemap",
  44. "description": "Orchestrate parallel codebase analysis to produce 7 structured documents (STACK, INTEGRATIONS, ARCHITECTURE, STRUCTURE, CONVENTIONS, TESTING, CONCERNS) in .planning/codebase/.",
  45. "version": "1.0.0",
  46. "author": {
  47. "name": "Dung Huynh Duc",
  48. "email": "dung@productsway.com"
  49. },
  50. "source": "./skills/codemap",
  51. "category": "productivity",
  52. "homepage": "https://github.com/jellydn/my-ai-tools"
  53. },
  54. {
  55. "name": "commit-atomic",
  56. "description": "Creates atomic commits by logically grouping staged changes with commitizen convention",
  57. "version": "1.0.0",
  58. "author": {
  59. "name": "Dung Huynh Duc",
  60. "email": "dung@productsway.com"
  61. },
  62. "source": "./skills/commit-atomic",
  63. "category": "productivity"
  64. },
  65. {
  66. "name": "docs-update",
  67. "description": "Update user-facing documentation when code changes. Use when asked to update docs, review docs, handle documentation changes, run scheduled documentation tasks, or analyze recent commits for documentation needs.",
  68. "version": "1.0.0",
  69. "author": {
  70. "name": "Dung Huynh Duc",
  71. "email": "dung@productsway.com"
  72. },
  73. "source": "./skills/docs-update",
  74. "category": "productivity"
  75. },
  76. {
  77. "name": "draft-pull-request",
  78. "description": "Creates or updates a draft pull request using gh CLI with a what/why/how description and optional change-outline views",
  79. "version": "1.0.0",
  80. "author": {
  81. "name": "Dung Huynh Duc",
  82. "email": "dung@productsway.com"
  83. },
  84. "source": "./skills/draft-pull-request",
  85. "category": "productivity"
  86. },
  87. {
  88. "name": "handoffs",
  89. "description": "Creates detailed handoff plans of conversations for continuing work in new sessions. Capture context, technical decisions, and pending tasks for seamless session continuation.",
  90. "version": "1.0.0",
  91. "author": {
  92. "name": "Dung Huynh Duc",
  93. "email": "dung@productsway.com"
  94. },
  95. "source": "./skills/handoffs",
  96. "category": "productivity"
  97. },
  98. {
  99. "name": "llm-wiki",
  100. "description": "Build and maintain a persistent, compounding knowledge wiki from raw sources. Use when the user wants to create a knowledge base, ingest documents into a wiki, query a wiki, health-check a wiki, or set up Karpathy's LLM Wiki pattern.",
  101. "version": "1.0.0",
  102. "author": {
  103. "name": "Dung Huynh Duc",
  104. "email": "dung@productsway.com"
  105. },
  106. "source": "./skills/llm-wiki",
  107. "category": "productivity",
  108. "homepage": "https://github.com/jellydn/my-ai-tools"
  109. },
  110. {
  111. "name": "pickup",
  112. "description": "Resumes work from previous handoff sessions stored in .planning/handoffs/. Load and continue from where you left off in prior sessions.",
  113. "version": "1.0.0",
  114. "author": {
  115. "name": "Dung Huynh Duc",
  116. "email": "dung@productsway.com"
  117. },
  118. "source": "./skills/pickup",
  119. "category": "productivity"
  120. },
  121. {
  122. "name": "plannotator-setup-goal",
  123. "description": "Turn an idea into a structured goal package via Plannotator-gated discovery, fact sheet, and plan.",
  124. "version": "1.0.0",
  125. "author": {
  126. "name": "Dung Huynh Duc",
  127. "email": "dung@productsway.com"
  128. },
  129. "source": "./skills/plannotator-setup-goal",
  130. "category": "productivity"
  131. },
  132. {
  133. "name": "portless-local",
  134. "description": "Named .localhost URLs for local development - replaces port numbers with stable, readable URLs",
  135. "version": "1.0.0",
  136. "author": {
  137. "name": "Dung Huynh Duc",
  138. "email": "dung@productsway.com"
  139. },
  140. "source": "./skills/portless-local",
  141. "category": "productivity"
  142. },
  143. {
  144. "name": "pr-review",
  145. "description": "Fix PR review comments by implementing requested changes. Fetch review comments, implement fixes, run tests, and commit changes.",
  146. "version": "1.0.0",
  147. "author": {
  148. "name": "Dung Huynh Duc",
  149. "email": "dung@productsway.com"
  150. },
  151. "source": "./skills/pr-review",
  152. "category": "productivity"
  153. },
  154. {
  155. "name": "prd",
  156. "description": "Generate Product Requirements Documents (PRD) for new features. Use when planning features, starting projects, or creating specifications.",
  157. "version": "1.0.0",
  158. "author": {
  159. "name": "Dung Huynh Duc",
  160. "email": "dung@productsway.com"
  161. },
  162. "source": "./skills/prd",
  163. "category": "productivity"
  164. },
  165. {
  166. "name": "qmd-knowledge",
  167. "description": "Project-specific knowledge management system using qmd MCP server. Captures learnings, issue notes, and conventions in a searchable knowledge base.",
  168. "version": "1.0.0",
  169. "author": {
  170. "name": "Dung Huynh Duc",
  171. "email": "dung@productsway.com"
  172. },
  173. "source": "./skills/qmd-knowledge",
  174. "category": "productivity",
  175. "homepage": "https://github.com/jellydn/my-ai-tools"
  176. },
  177. {
  178. "name": "ralph",
  179. "description": "Convert PRDs to prd.json format for the Ralph autonomous agent system. Converts markdown PRDs to Ralph's JSON format.",
  180. "version": "1.0.0",
  181. "author": {
  182. "name": "Dung Huynh Duc",
  183. "email": "dung@productsway.com"
  184. },
  185. "source": "./skills/ralph",
  186. "category": "productivity"
  187. },
  188. {
  189. "name": "security-audit",
  190. "description": "Perform structured security audits using OWASP Top 10, ASVS, cheat sheets, framework checks, and DevSecOps controls.",
  191. "version": "1.0.0",
  192. "author": {
  193. "name": "Dung Huynh Duc",
  194. "email": "dung@productsway.com"
  195. },
  196. "source": "./skills/security-audit",
  197. "category": "code-quality"
  198. },
  199. {
  200. "name": "slop",
  201. "description": "Removes AI-generated code slop from git diffs to maintain code quality. Clean up extra comments, defensive checks, type casts, and inconsistent style.",
  202. "version": "1.0.0",
  203. "author": {
  204. "name": "Dung Huynh Duc",
  205. "email": "dung@productsway.com"
  206. },
  207. "source": "./skills/slop",
  208. "category": "productivity"
  209. },
  210. {
  211. "name": "tdd",
  212. "description": "Guides through the complete TDD workflow with Red-Green-Refactor cycle. Start features, write failing tests, implement minimal code, and refactor with confidence.",
  213. "version": "1.0.0",
  214. "author": {
  215. "name": "Dung Huynh Duc",
  216. "email": "dung@productsway.com"
  217. },
  218. "source": "./skills/tdd",
  219. "category": "productivity"
  220. },
  221. {
  222. "name": "code-quality-review",
  223. "description": "Run an extremely strict maintainability and structural code quality review — flags abstraction issues, spaghetti growth, and boundary leaks",
  224. "version": "1.0.0",
  225. "author": {
  226. "name": "Dung Huynh Duc",
  227. "email": "dung@productsway.com"
  228. },
  229. "source": "./skills/code-quality-review",
  230. "category": "code-quality"
  231. },
  232. {
  233. "name": "code-review",
  234. "description": "Review the diff since a fixed point along two axes — Conventions (does the code follow this repo's coding standards and Tidy First practices?) and Intent (does the change do what it claims to do?). Runs both reviews in parallel sub-agents and reports them side by side.",
  235. "version": "1.0.0",
  236. "author": {
  237. "name": "Dung Huynh Duc",
  238. "email": "dung@productsway.com"
  239. },
  240. "source": "./skills/code-review",
  241. "category": "code-quality"
  242. },
  243. {
  244. "name": "tmux",
  245. "description": "Remote control tmux sessions for interactive CLIs (python, node, gdb, etc.) by sending keystrokes and capturing pane output. Integrates with LogPilot for AI-assisted log analysis.",
  246. "version": "1.0.0",
  247. "author": {
  248. "name": "Dung Huynh Duc",
  249. "email": "dung@productsway.com"
  250. },
  251. "source": "./skills/tmux",
  252. "category": "terminal-automation",
  253. "homepage": "https://github.com/jellydn/logpilot"
  254. },
  255. {
  256. "name": "visual-pr",
  257. "description": "Posts a concise visual outline as a GitHub pull request comment without replacing the pull request description.",
  258. "version": "1.0.0",
  259. "author": {
  260. "name": "Dung Huynh Duc",
  261. "email": "dung@productsway.com"
  262. },
  263. "source": "./skills/visual-pr",
  264. "category": "productivity"
  265. }
  266. ]
  267. }
README.md · README · 509 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. # Welcome to my-ai-tools 👋
  2. [![Tests](https://github.com/jellydn/my-ai-tools/actions/workflows/test.yml/badge.svg)](https://github.com/jellydn/my-ai-tools/actions/workflows/test.yml)
  3. [![GitHub stars](https://img.shields.io/github/stars/jellydn/my-ai-tools)](https://github.com/jellydn/my-ai-tools/stargazers)
  4. [![GitHub license](https://img.shields.io/github/license/jellydn/my-ai-tools)](LICENSE)
  5. [![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg)](CONTRIBUTING.md)
  6. > Portable, source-controlled configuration for AI coding assistants.
  7. `my-ai-tools` installs one consistent set of agent guidelines, skills, commands, hooks, MCP servers, and tool
  8. settings across supported AI coding assistants. It can also export local configuration changes back into this
  9. repository.
  10. 📖 [Documentation website](https://ai-tools.itman.fyi) · 🎬 [Setup video](https://www.youtube.com/watch?v=ESudSFAyuuw)
  11. ## ✨ Features
  12. - Configures Claude Code, OpenCode, Amp, Codex, Cursor, Pi, Antigravity, and many other assistants.
  13. - Shares reusable skills, agent guidance, commands, hooks, and MCP server definitions.
  14. - Supports macOS, Linux, Windows 11 with PowerShell and Git Bash, and WSL.
  15. - Provides side-effect-free dry runs and optional backups before installation.
  16. - Exports installed configurations back to the repository for bidirectional sync.
  17. - Validates JSON and other supported configuration files before installation.
  18. ## 📋 Requirements
  19. ### macOS, Linux, and WSL
  20. - Bash 3.0 or later
  21. - Git
  22. - `curl` for the one-line installer
  23. - Bun or Node.js LTS for JavaScript-based tools
  24. ### Windows 11
  25. - Windows PowerShell 5.1 or PowerShell 7
  26. - [`winget`](https://learn.microsoft.com/windows/package-manager/winget/) (App Installer) for automatic prerequisite setup
  27. The installer handles Git for Windows (including Git Bash), `jq`, and Node.js LTS with npm. You do not need to
  28. install them separately or open Git Bash first. WSL users can use the Linux instructions.
  29. ## 🚀 Installation
  30. The installer can replace existing configuration files. Run a dry run first and use `--backup` or `-Backup` when
  31. you want a restorable copy.
  32. ### macOS, Linux, and WSL
  33. Review the installer, then run it:
  34. ```bash
  35. curl -fsSL https://ai-tools.itman.fyi/install.sh -o install.sh
  36. less install.sh
  37. bash install.sh --dry-run
  38. bash install.sh --backup
  39. ```
  40. For a direct one-line installation:
  41. ```bash
  42. curl -fsSL https://ai-tools.itman.fyi/install.sh | bash
  43. ```
  44. ### Windows 11
  45. Open PowerShell and run the canonical installer:
  46. ```powershell
  47. Invoke-RestMethod https://ai-tools.itman.fyi/install.ps1 | Invoke-Expression
  48. ```
  49. This downloads and executes a remote script. Review [install.ps1](install.ps1) before trusting it. It installs
  50. missing prerequisites through `winget`, reuses existing tools, refreshes PATH for the current process, and starts
  51. the shared installer in Git Bash. Windows may request administrator approval for Git or Node.js installation.
  52. Package installation uses `--source winget`, not the Microsoft Store. Node and npm are discovered again after
  53. installation and passed to Git Bash in the same process; no separate prerequisite command or terminal restart
  54. is needed after a successful install. The wrapper skips Git Bash login profiles so an older shell PATH cannot
  55. hide these tools. Approve only the package installer's elevation prompt; the wrapper does not relaunch the full
  56. configuration installer as administrator. If approval is denied, setup stops and can be rerun safely.
  57. Native Windows uses Node/npm, not Bun, for npm package installation; Linux, macOS, and WSL keep Bun-first behavior.
  58. Amp is not installed through npm. Setup uses the [official Amp installers](https://ampcode.com/docs/cli):
  59. `install.ps1` on native Windows (including Git Bash), and `install.sh` on macOS, Linux, and WSL.
  60. Setup checks `amp version` and only reuses the binary in `AMP_HOME/bin` (default: `~/.amp/bin`);
  61. finding another command on PATH is not enough, including an npm shim under a custom prefix.
  62. Symlinks to the official binary are recognised and do not trigger an npm warning.
  63. An exported Bash `amp()` function produces a warning; setup verifies the official binary directly without removing or unsetting the function.
  64. `AMP_HOME` may be a Windows path or a Git Bash path; setup converts it for each side, so the
  65. official installer and the verification always use the same directory.
  66. Our wrapper promotes the official bin directory in the current Bash and PowerShell processes only.
  67. The official Windows `install.ps1` adds `AMP_HOME\bin` (default: `.amp\bin`) to the saved User PATH when absent;
  68. it does not reorder an existing entry. The official Unix `install.sh` can add `~/.local/bin` to a shell profile
  69. when needed. These official-installer changes persist across sessions; the wrapper's promotion does not.
  70. The PowerShell wrapper checks Amp only if it was accepted in setup; declining Amp leaves other tool installs unaffected.
  71. The one-line command and `-y` / `--yes` use these default categories:
  72. - **Default tools:** Amp, Pi, Codex, OpenCode, Claude Code, Grok, qmd, rtk, ctx, open-cursor, ai-switcher, open_code_review, and Plannotator.
  73. - **Shared skills:** bundled local skills, installed for the selected agents.
  74. - **Applicable MCP/plugin configuration:** bundled configuration for the selected agents. The three memory
  75. MCP choices (qmd, codebase-memory-mcp, agentmemory) remain available; `user-memory` stays opt-in.
  76. - **Optional standalone catalog:** remaining CLIs, formatters, and tools need selection. Interactive runs
  77. prompt for them; automation can use `--with-tools`.
  78. qmd, ctx, and Plannotator are also logged as integration dependencies of the core agent templates.
  79. ctx installation in `-y` mode does not import history, start its daemon, install upstream skills, or change
  80. persistent Windows PATH. Other MCP commands keep their existing external prerequisites. No unrelated agent's
  81. plugin or skill directory is configured in yes mode.
  82. From a reviewed checkout, select extra tools without enabling integrations:
  83. ```bash
  84. ./cli.sh --yes --without-integrations --with-tools=fx,cursor --dry-run
  85. ./cli.sh --yes --without-integrations --with-tools=fx,cursor
  86. # Default configuration, plugins, shared skills and MCPs (can replace existing config):
  87. ./cli.sh --yes --backup
  88. ```
  89. Windows equivalents are `-WithoutIntegrations -WithTools 'fx,cursor'` and `-Yes -Backup` on a saved
  90. `install.ps1`. Tool names are comma-separated; `./cli.sh --help` lists the catalog. `--without-integrations`
  91. (`-WithoutIntegrations`) installs standalone CLIs only and leaves existing configuration unchanged.
  92. `--with-integrations` remains accepted but is now the default. Selecting an extra tool does not select
  93. another agent's configuration. Noninteractive/CI runs use the same default categories as `--yes`.
  94. The installer can replace existing configuration. To inspect the script, preview changes, and request a backup,
  95. use the saved-script path instead:
  96. ```powershell
  97. Invoke-RestMethod https://ai-tools.itman.fyi/install.ps1 -OutFile install.ps1
  98. Get-Content .\install.ps1
  99. powershell -NoProfile -ExecutionPolicy Bypass -File .\install.ps1 -DryRun
  100. powershell -NoProfile -ExecutionPolicy Bypass -File .\install.ps1 -Backup
  101. ```
  102. Execution-policy bypass applies only to that process. `-DryRun` does not install prerequisites; install them first
  103. if the preview reports they are missing. The one-line command is non-interactive and automatically uses `--yes`.
  104. If setup fails, resolve the reported error and rerun the same command. If a newly installed prerequisite still
  105. cannot be found, close all terminal windows and reopen PowerShell. If `winget` is missing, install or update
  106. [App Installer](https://aka.ms/getwinget), or install [Git for Windows](https://git-scm.com/download/win),
  107. [`jq`](https://jqlang.org/download/), and [Node.js LTS with npm](https://nodejs.org/) manually. In Git setup,
  108. select **Git from the command line and also from 3rd-party software**. Do not bypass package lifecycle errors
  109. with `--ignore-scripts`: native binaries can remain uninstalled.
  110. #### Windows recovery: Amp command priority
  111. If `amp` reports an incompatible npm build, or PowerShell blocks `AppData\Roaming\npm\amp.ps1`,
  112. check command selection before changing anything. The wrapper verifies the official binary, puts its directory
  113. first in the current PowerShell process PATH, and checks command selection again. A shadowing alias, function,
  114. or a PATH entry that a new terminal would pick first is reported as a warning and does not fail the rest of
  115. the setup; only a missing or broken official binary fails the run.
  116. ```powershell
  117. Get-Command amp -All
  118. $ampHome = if ($env:AMP_HOME) { $env:AMP_HOME } else { Join-Path $env:USERPROFILE '.amp' }
  119. $ampBin = Join-Path $ampHome 'bin'
  120. & (Join-Path $ampBin 'amp.exe') version
  121. # Current session only: prefer the official binary, without removing npm or other tools.
  122. $env:Path = "$ampBin;$env:Path"
  123. Get-Command amp -All
  124. amp version
  125. ```
  126. For future sessions, open Windows **Environment Variables** and place the official Amp bin directory before
  127. `AppData\Roaming\npm` in PATH, then close and reopen PowerShell. Do not replace the full PATH or remove npm's
  128. directory: other tools can need it. Review any `amp` alias or function separately; PATH does not override those.
  129. If the official binary is missing, install it with
  130. `Invoke-RestMethod https://ampcode.com/install.ps1 | Invoke-Expression`, then repeat the checks above.
  131. Do not reinstall `@ampcode/cli` through npm or lower execution policy to run its stale shim. Setup does not
  132. uninstall existing copies, delete shims, or change execution policy. A direct binary version failure needs
  133. investigation; changing PATH alone cannot repair an incompatible binary.
  134. In Bash or Git Bash, run `type amp` in your own shell to check for a function. Unexported functions from
  135. shell profiles are not visible to the child Bash process running setup.
  136. Claude is skipped only when `claude --version` succeeds with nonempty output. A broken shim (including a
  137. missing native binary) is repaired even with `--yes`, then checked again. npm installs for Claude and OpenCode
  138. explicitly include optional platform packages and enable lifecycle scripts for that command only; Claude
  139. repair also uses `--force`. When npm supports script approval, Claude installation adds
  140. `--allow-scripts=@anthropic-ai/claude-code` to permit only that package's lifecycle scripts for that invocation.
  141. Older npm versions without this setting omit the flag. These options do not change user/global npm settings
  142. or approve scripts for other packages. Installation still fails if the final `claude --version` check fails.
  143. If OpenCode postinstall reports `ECONNRESET`, its platform download failed; the log alone does not identify
  144. the network cause. Check registry connectivity and your network/proxy policy, then retry
  145. `npm install -g @opencode/cli --include=optional --ignore-scripts=false`, verify `opencode --version`, and rerun
  146. setup. Optional catalog tools, including OpenCode 2, AI Launcher (`ai-switcher`), and Pi, warn and continue
  147. after installation or verification fails. Setup lists failed/skipped optional components at the end and
  148. skips their configurations for that run. Rerun setup to retry them. Claude, Amp, Codex, prerequisite checks,
  149. and required integration dependencies (including qmd, ctx, and Plannotator) still fail setup with a nonzero
  150. exit code. Optional failure does not mean that tool was installed successfully.
  151. AI Launcher reports success only after its installer succeeds and `ai --version` returns nonempty output
  152. successfully. On native Windows, setup downloads the official release directly instead of executing the
  153. upstream installer, which has no download retries and writes directly to the installed binary. The new
  154. release is staged beside the destination, checked against the release's SHA-256 checksum, and run with
  155. `--version` before it replaces `ai.exe`. Failed downloads, checksum checks, or staged health checks preserve
  156. the existing binary. A missing Windows architecture asset is reported, not substituted with another build.
  157. Shared installer downloads make at most three GET attempts for curl connection/partial-transfer/timeout/
  158. empty-response/receive errors (7, 18, 28, 52, 56), with 1-second then 2-second waits. Curl 35 retries only
  159. when that attempt's stderr matches `connection (was )?reset`, ignoring case. Generic TLS handshake errors,
  160. HTTP errors, and TLS certificate failures do not retry. Earlier attempts and stdout cannot enable a retry.
  161. Each attempt has a 15-second connection limit and 120-second transfer limit. Persistent TLS resets stop
  162. after three attempts; retries do not identify the cause or repair certificate, protocol, or proxy problems.
  163. TLS verification remains enabled on every attempt.
  164. Partial files are restarted, not appended; checksum verification still applies. Downloaded scripts
  165. and npm installs are never automatically replayed. This can recover a transient interruption, not repair
  166. a persistent network problem. Unix/WSL AI Launcher installation still uses the upstream installer.
  167. #### Windows recovery: Pi and persistent download resets
  168. npm already retries eligible registry reads (default `fetch-retries=2`, delays of 10 to 60 seconds, and a
  169. 5-minute per-request timeout). Local settings can differ. Setup does not add an outer Pi install retry or
  170. change user/global npm settings. Check the configured registry and effective retry count first:
  171. ```powershell copy Copy Registry Checks
  172. node --version
  173. npm --version
  174. npm config get registry
  175. npm config get fetch-retries
  176. npm ping
  177. ```
  178. `npm ping` tests registry metadata access only, not every package tarball or GitHub release download.
  179. Pi's `EPERM` cleanup warnings do not establish why Windows refused a removal. Save the npm debug log,
  180. close any Pi process you started, and use Resource Monitor or Process Explorer to check the exact failing
  181. path for an open handle. Check that directory's permissions if no handle is found. Do not stop all Node
  182. processes, run blanket deletion/cache cleanup, disable TLS checks, or use administrator mode as a workaround.
  183. After network access and any confirmed file-access problem are resolved, retry once from PowerShell:
  184. ```powershell copy Copy Pi Recovery
  185. npm install -g @earendil-works/pi-coding-agent --loglevel=http --timing
  186. pi --version
  187. ```
  188. Run the health check even if npm reports success. If failures persist, share the **redacted npm debug log**
  189. with failed URL host/path, HTTP status or socket error, attempt counts, timestamps, and Node/npm versions.
  190. For AI Launcher, include Git Bash's `curl --version` and a redacted verbose trace of the failing release
  191. asset GET, including redirects and response headers. Say whether a proxy/VPN is in use and whether another
  192. approved network reproduces it. Remove tokens, cookies, proxy credentials, and signed URL query strings.
  193. For `EPERM`, include the exact affected path, the confirmed handle/permission findings, and whether the error
  194. also occurs during the isolated Pi retry. These are needed to separate local file access from network resets.
  195. Optional qmd requires Node.js >=22 on native Windows. Its installer prioritizes the real `node.exe` directory
  196. and npm's global prefix in the same Git Bash session, ahead of stale Bun launchers or extensionless Node copies
  197. in `~/.local/bin`. Those old files are not deleted. In npm 11.19.1, unreviewed `allowScripts` entries warn but
  198. still run unless strict policy is enabled. The installer does not change npm script policy. Review actual
  199. build errors and approve only required packages if your policy requires it; do not enable all scripts globally.
  200. On Windows ARM64, qmd keyword indexing/search works, but sqlite-vec currently warns that vector search is unsupported.
  201. Optional Plannotator installs only the official v0.29.0 Windows release binary (x64 or ARM64), with a pinned
  202. SHA-256 hash. It uses `%LOCALAPPDATA%\plannotator` and updates only the current Git Bash PATH. No remote
  203. PowerShell script, policy bypass, persistent PATH change, sidecar, skill, or agent-settings edit is used.
  204. The direct method avoids an upstream PowerShell `$Minimal`/`$minimal` variable collision. Failure output gives
  205. the release URL, expected hash, and destination for manual recovery. In a separate terminal, add that directory
  206. and npm's global prefix to your user PATH, or use full executable paths. Unix/WSL installation paths are unchanged.
  207. #### Plannotator installation and Inbox CLI
  208. On macOS, Linux, and WSL, install Plannotator with the official installer:
  209. ```bash
  210. curl -fsSL https://plannotator.ai/install.sh | bash
  211. ```
  212. The default my-ai-tools installation runs this same installer on Unix-like systems. On native Windows,
  213. it uses the pinned official v0.29.0 binary fallback documented above instead of executing a remote PowerShell script.
  214. After installation, use the `plannotator inbox` command to open and manage the local Inbox:
  215. ```bash
  216. # Start the Inbox and open it in your browser
  217. plannotator inbox
  218. # Open the running Inbox and print its URL without blocking
  219. plannotator inbox --open
  220. # Start it in the background and print the URL without opening a browser
  221. plannotator inbox --background
  222. # Check whether it is running, including its URL and data directory
  223. plannotator inbox status
  224. # Run the stdio MCP server for an agent that has no native Inbox integration
  225. plannotator inbox mcp
  226. ```
  227. The installer configures `plannotator inbox mcp` for supported MCP clients. For agents with a native
  228. Inbox integration, use the agent's `plannotator_inbox` tool when available. The Inbox tool is enabled for
  229. new sessions after the Inbox has been started once; restart the agent to load it in an existing session.
  230. Use `plannotator inbox stop` only when you want to stop the shared Inbox, because other agent sessions use
  231. the same process. Run `plannotator inbox --help` for the complete CLI reference.
  232. If the wrapper says all prerequisites are met but Bash then reports missing Node/npm, check that the downloaded
  233. `install.ps1` includes `Find-Node` and the `OpenJS.NodeJS.LTS` installation step. A stale web deployment can serve
  234. an older wrapper while cloning the latest Bash installer. Maintainers must check the **Dokku Deploy** workflow
  235. for `ai-tools.itman.fyi`; a successful **Deploy GitHub Pages** run does not update that domain. Until the domain
  236. is updated, use a reviewed repository checkout and run:
  237. ```powershell
  238. powershell -NoProfile -ExecutionPolicy Bypass -File .\install.ps1 -Yes
  239. ```
  240. ### Repository checkout
  241. ```bash
  242. git clone https://github.com/jellydn/my-ai-tools.git
  243. cd my-ai-tools
  244. ./cli.sh --dry-run
  245. ./cli.sh --backup
  246. ```
  247. On Windows PowerShell, use the native wrapper from the checkout:
  248. ```powershell
  249. git clone https://github.com/jellydn/my-ai-tools.git
  250. Set-Location my-ai-tools
  251. .\install.ps1 -DryRun
  252. .\install.ps1 -Backup
  253. ```
  254. ## ⚙️ Installer options
  255. | Bash option | PowerShell option | Purpose |
  256. | --- | --- | --- |
  257. | `--dry-run` | `-DryRun` | Show planned changes without applying them. |
  258. | `--backup` | `-Backup` | Back up existing files before installation. |
  259. | `--no-backup` | `-NoBackup` | Skip the interactive backup prompt. |
  260. | `--yes` / `-y` | `-Yes` | Install the default tool set, shared skills, applicable MCP/plugin config and declared dependencies. |
  261. | `--with-tools=fx,cursor` | `-WithTools 'fx,cursor'` | Add optional standalone catalog CLIs. |
  262. | `--with-integrations` | `-WithIntegrations` | Install bundled config, plugins, skills and MCP dependencies (default). |
  263. | `--without-integrations` | `-WithoutIntegrations` | Install standalone CLIs only; leave configuration unchanged. |
  264. | `--rollback` | `-Rollback` | Restore the most recent backup. |
  265. | `--verbose` / `-v` | `-Verbose` | Show detailed installer output. |
  266. | `--migrate-gemini` | — | Migrate Gemini CLI configuration to Antigravity CLI. |
  267. Non-interactive Bash and PowerShell pipelines automatically add `--yes`. This mode uses the same default
  268. categories above; it is not a configuration-free install. Review
  269. [`cli.sh`](cli.sh) before using it in automation.
  270. ## 🤖 Supported tools
  271. The repository currently contains configuration or integration support for:
  272. | Category | Tools |
  273. | --- | --- |
  274. | Primary assistants | Claude Code, OpenCode, Amp, Codex, Cursor, Pi, Oh My Pi, Antigravity, Kilo, Kimi Code, Muse Code |
  275. | Additional CLIs | CommandCode, GitHub Copilot CLI, Gemini CLI, Grok CLI, MiMo-Code, Qoder CLI, DeepSeek Harness, Kiro CLI, Devin CLI, Factory Droid, Cline, Reasonix |
  276. | Workflow and desktop tools | Conductor, Delta, Codiff, OpenMausBot, Hunk, ctx, herdr, Orca, AI Launcher, CCS, fx, Open Code Review, ghui |
  277. Some installers are platform-specific, and optional tools are skipped when they are not detected. Gemini CLI is
  278. deprecated for Google One and unpaid tiers; use the `--migrate-gemini` option to move its configuration to
  279. Antigravity CLI.
  280. See [`configs/`](configs/) for the source configuration of each tool and [`skills/`](skills/) for shared skills.
  281. For an opt-in VPS terminal accessed from a Mac over Tailscale, see the
  282. [Pi Durable setup](configs/pi/durable/README.md). It uses a pinned experimental
  283. TUI, private persistent storage, and a user service; the general installer does not enable it.
  284. ## 🔌 Shared configuration
  285. The installer manages these common resources where the target tool supports them:
  286. - Agent guidance and software-development practices
  287. - Reusable skills and commands
  288. - Lifecycle and safety hooks
  289. - Tool-specific settings, themes, and provider configuration
  290. - MCP servers from [`configs/mcp-registry.json`](configs/mcp-registry.json)
  291. The MCP registry includes integrations such as Context7, sequential thinking, qmd, codebase memory, agent memory,
  292. fff, React Grab, logpilot, sem, and ctx. Installation depends on each server's runtime and credentials.
  293. For explicit cross-project user preferences, [user-memory MCP](packages/user-memory-mcp/README.md) provides
  294. five preference tools backed by local files. It is **optional**: the installer does not install or configure it.
  295. The existing qmd, codebase-memory-mcp, and agentmemory integrations remain unchanged. Follow the package guide
  296. to build, link, and register user-memory only if you need a separate explicit-preference store.
  297. See [the memory guide](MEMORY.md) for the distinction between preferences, project knowledge, and session handoffs.
  298. ctx's initial history indexing is optional. If provider paths cannot be inspected safely, the installer retains
  299. the verified binary and continues with a warning. Retry with `ctx setup`; see [ctx troubleshooting](TOOL_REFERENCE.md#-ctx-optional)
  300. for exact-path imports without bypassing its safety checks.
  301. Do not commit secrets. Keep API keys in environment variables or the target tool's credential store. Review
  302. `.env.example` for server-side environment variable names used by this repository.
  303. ## 🔄 Export local changes
  304. `generate.sh` copies supported configuration from your home directory back into the repository. Preview the
  305. operation first:
  306. ```bash
  307. ./generate.sh --dry-run
  308. ./generate.sh
  309. git diff
  310. ```
  311. Review the diff before committing. Generated configuration can contain machine-specific values or credentials that
  312. must not enter version control.
  313. ## 🧭 Common workflows
  314. ### Restore a backup
  315. ```bash
  316. ./cli.sh --rollback
  317. ```
  318. ```powershell
  319. .\install.ps1 -Rollback
  320. ```
  321. Backups are stored in `~/ai-tools-backup-<timestamp>`, and the installer keeps the five most recent backups.
  322. ### Migrate Gemini CLI to Antigravity CLI
  323. ```bash
  324. ./cli.sh --migrate-gemini
  325. ```
  326. ### Validate a contribution
  327. ```bash
  328. bash -n cli.sh generate.sh install.sh lib/*.sh scripts/*.sh
  329. ./cli.sh --dry-run
  330. ./scripts/sync-token-efficiency.sh --check
  331. bats tests/
  332. biome check .
  333. ```
  334. See [`TESTING.md`](TESTING.md) for focused test commands and [`CONTRIBUTING.md`](CONTRIBUTING.md) for repository
  335. conventions.
  336. ### Update shared agent guidelines
  337. Edit [`configs/token-efficiency.md`](configs/token-efficiency.md) or
  338. [`configs/experiment-workspaces.md`](configs/experiment-workspaces.md), then run:
  339. ```bash
  340. ./scripts/sync-token-efficiency.sh
  341. ./scripts/sync-token-efficiency.sh --check
  342. ```
  343. The script syncs both sections across the managed profiles. Experiment Workspaces also covers the Claude Code
  344. profile and the root `AGENTS.md` and `GEMINI.md`. Do not edit these generated sections in individual profiles;
  345. the CI check rejects drift from the canonical files. The workspace guideline uses `try` when available, but
  346. the installer does not install it. It requires shell integration; fresh experiments use an interactive selector,
  347. with fallbacks for agent shells without a terminal. `generate.sh` restores the canonical sections after export,
  348. so an older home-directory profile cannot remove them.
  349. ## 🎨 GitHub Coding Taste Generator
  350. `code-taste` analyzes public GitHub repositories and generates an evidence-backed
  351. coding preference profile. This command requires Bun and an OpenAI-compatible API
  352. key. Tree-sitter extracts TypeScript/TSX declarations; Markdown chunks retain headings
  353. and fenced code. See [setup and model configuration](docs/code-taste-openrouter.md).
  354. ```bash
  355. bun install
  356. bun run code-taste analyze jellydn/my-ai-tools
  357. bun run code-taste analyze jellydn --repos 3 --language TypeScript
  358. bun run code-taste export --format markdown
  359. ```
  360. Analysis state is stored in ignored `.code-taste/`. `CODING_TASTE.md` is intentionally
  361. not ignored so you can review and commit the resulting guidance. Markdown omits evidence;
  362. review the evidence in `.code-taste/profile.json` or the JSON export before treating
  363. inferred preferences as project rules.
  364. ## 📚 Documentation
  365. HTML pages in [`docs/`](docs/index.html) are generated from the Markdown sources. Regenerate them with `bun scripts/render-docs.ts`.
  366. - [Documentation website](https://ai-tools.itman.fyi)
  367. - [Detailed tool and skill reference](TOOL_REFERENCE.md)
  368. - [Fusion orchestration](docs/fusion-orchestration.md)
  369. - [Claude Code agent teams](docs/claude-code-teams.md)
  370. - [30-day applied AI learning sprint](docs/30-day-applied-ai-learnings.md)
  371. - [qmd knowledge management](docs/qmd-knowledge-management.md)
  372. - [My AI Bot](docs/my-ai-bot.md)
  373. - [Public repository knowledge base](docs/rag-architecture.md)
  374. - [Retrieval evaluation](docs/rag-retrieval-quality.md)
  375. - [Applied AI and GenAI cheat sheet](docs/ai-applied-genai-cheat-sheet.md)
  376. - [Testing guide](TESTING.md)
  377. ## 🤝 Contributing
  378. Contributions, issues, and feature requests are welcome. Read [`CONTRIBUTING.md`](CONTRIBUTING.md), run the relevant
  379. checks, and open an issue before a large behavior or configuration change.
  380. ## 📜 License
  381. This project is available under the [MIT License](LICENSE).
  382. ## 👤 Author
  383. **Dung Huynh**
  384. - Website: [productsway.com](https://productsway.com/)
  385. - YouTube: [IT Man Channel](https://www.youtube.com/@it-man)
  386. - X: [@jellydn](https://twitter.com/jellydn)
  387. - GitHub: [@jellydn](https://github.com/jellydn)
  388. ## ⭐ Show your support
  389. Give this project a ⭐️ if it helps you.
  390. [![Ko-fi](https://img.shields.io/badge/Ko--fi-Support-ff5e5b?logo=ko-fi&logoColor=white)](https://ko-fi.com/dunghd)
  391. [![PayPal](https://img.shields.io/badge/PayPal-Support-00457C?logo=paypal&logoColor=white)](https://paypal.me/dunghd)
  392. [![Buy Me a Coffee](https://img.shields.io/badge/Buy_Me_a_Coffee-Support-FFDD00?logo=buymeacoffee&logoColor=000000)](https://www.buymeacoffee.com/dunghd)

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.