AI-Skills

Personal AI skills for any coding agent — install once, use everywhere.

Quick Install

macOS / Linux / Git Bash:

curl -fsSL https://raw.githubusercontent.com/chanakya-net/AI-Skills/main/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/chanakya-net/AI-Skills/main/install.ps1 | iex

The smart installer detects your active agent(s) and wires everything up automatically. It also installs shared assets required by workflow skills:

• prompt.md
• run-agent.sh
• run-agent.ps1 on Windows
• agent-registry.json

Default asset location:

macOS/Linux:  ~/.ai-skill-collections/assets
Windows:      %USERPROFILE%\.ai-skill-collections\assets

Runner contract (execution-only):

Bash (macOS / Linux / Git Bash):

• run-agent.sh --agent <agent> --context-file <context-payload-file> --prompt-file <prompt-file>

PowerShell (Windows):

• run-agent.ps1 --agent <agent> --context-file <context-payload-file> --prompt-file <prompt-file>

run-with-it prepares context and invokes the runner. The runner does not fetch GitHub issues or git history itself.

Codebase Overview

This repository packages a small skill collection plus the shared runner assets those skills need.

The codebase has four main surfaces:

• skills/ contains the agent-facing skill instructions. Each skill is a standalone SKILL.md with YAML front matter.
• assets/ contains runtime assets used by run-with-it, especially the unified runner scripts and agent/model registry.
• install.sh / install.ps1 detect local agent tools and install the collection for each supported environment.
• uninstall.sh / uninstall.ps1 remove agent registrations, shared assets, and this collection's installed skill directories for a clean reinstall.

The repository intentionally keeps the runtime contract simple: skills prepare context, then run-with-it selects an agent/model and invokes assets/run-agent.sh or assets/run-agent.ps1. The runner executes a prepared payload; it does not create issues, fetch issue bodies, or infer project history on its own.

Unified Routing Workflow

Use this sequence for issue-driven execution:

1. break-req captures and resolves requirements/constraints.
2. create-git-issue publishes prd.md + implementation slices with routing hints.
3. run-with-it performs final runtime routing and invokes run-agent.sh.

create-git-issue hints are advisory only. run-with-it remains final routing authority at execution time.

Routing and Registry Overrides

run-with-it uses agent-registry.json + complexity scoring to select agent/model. Automatic routing reserves Google/Gemini models for last-resort fallback only. Explicit AGENT or MODEL overrides can still select Gemini after normal validation.

Supported overrides and filters:

• AGENT_REGISTRY_FILE override registry path (default: <asset-root>/agent-registry.json)
• AGENT_ALLOWLIST comma-separated agent slugs to permit
• AGENT_DENYLIST comma-separated agent slugs to block (denylist wins conflicts)
• AGENT force agent selection (must be installed/valid)
• MODEL force model selection for selected agent
• COMPLEXITY_LEVEL force complexity band
• COMPLEXITY_SCORE force numeric score (8-40)

Override asset destination or git ref:

Bash:

ASSETS_DEST="$HOME/.my-ai-assets" ASSETS_REF=main curl -fsSL https://raw.githubusercontent.com/chanakya-net/AI-Skills/main/install.sh | bash

PowerShell:

$env:ASSETS_DEST="$env:USERPROFILE\.my-ai-assets"; $env:ASSETS_REF="main"
irm https://raw.githubusercontent.com/chanakya-net/AI-Skills/main/install.ps1 | iex

Troubleshooting

"No asset files" when running run-with-it

This means the shared files were not found in either:

• ~/.ai-skill-collections/assets (macOS/Linux) or %USERPROFILE%\.ai-skill-collections\assets (Windows)
• ./assets (current working directory)

Quick fix from this repository root:

Bash (macOS / Linux / Git Bash):

mkdir -p "$HOME/.ai-skill-collections/assets" && cp -f ./assets/prompt.md ./assets/run-agent.sh ./assets/agent-registry.json "$HOME/.ai-skill-collections/assets/" && chmod +x "$HOME/.ai-skill-collections/assets/run-agent.sh"

PowerShell (Windows):

New-Item -ItemType Directory -Force "$env:USERPROFILE\.ai-skill-collections\assets"; Copy-Item -Force .\assets\prompt.md, .\assets\run-agent.ps1, .\assets\run-agent.sh, .\assets\agent-registry.json "$env:USERPROFILE\.ai-skill-collections\assets\"

Or re-run installer:

Bash: bash install.sh PowerShell: .\install.ps1

No git repo yet

run-with-it can still run without git initialization. It will skip commit-history context and continue with issue/local context.

---

Uninstall

Run the full uninstaller to remove skills per agent and delete shared assets. This leaves the machine ready for a clean install.

macOS / Linux / Git Bash:

curl -fsSL https://raw.githubusercontent.com/chanakya-net/AI-Skills/main/uninstall.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/chanakya-net/AI-Skills/main/uninstall.ps1 | iex

Fresh reinstall from this repository root:

Bash:

bash uninstall.sh && bash install.sh

PowerShell:

.\uninstall.ps1; .\install.ps1

The uninstaller deletes shared assets by default:

macOS/Linux:  ~/.ai-skill-collections
Windows:      %USERPROFILE%\.ai-skill-collections

It also removes this collection's installed skill directories from standard skill roots such as ~/.agents/skills, ~/.codex/skills, and ~/.Codex/skills.

To preview removals:

Bash:

bash uninstall.sh --dry-run

PowerShell:

.\uninstall.ps1 -DryRun

Manual commands are also available if you want to remove one agent by hand.

Claude Code

claude plugin uninstall ai-skill-collections

Gemini CLI

gemini extensions uninstall https://github.com/chanakya-net/AI-Skills

Codex / Copilot / Antigravity (via npx)

npx -y skills remove chanakya-net/AI-Skills --global

Delete shared assets

macOS / Linux:

rm -rf "$HOME/.ai-skill-collections"

Windows (PowerShell):

Remove-Item -Recurse -Force "$env:USERPROFILE\.ai-skill-collections"

---

Per-Agent Install

Per-agent commands below install skills for that specific agent. To guarantee shared assets are installed too, prefer the install.sh one-liner above.

Claude Code

claude plugin install github:chanakya-net/AI-Skills

Gemini CLI

gemini extensions install github.com/chanakya-net/AI-Skills

All Other Agents (via npx)

npx -y skills add chanakya-net/AI-Skills -a <agent>

Agent	--agent slug
OpenAI Codex (CLI + GUI)	codex
GitHub Copilot (CLI + VS Code)	github-copilot
Gemini GUI (Antigravity)	antigravity

OpenCode note: OpenCode users must configure their preferred model in their local OpenCode setup; this repo does not set OpenCode model defaults for you.

Example:

npx -y skills add chanakya-net/AI-Skills -a github-copilot

---

Skills

Skill	Description
save-tokens	Compresses AI responses using symbols/abbreviations to cut token usage ~75%
break-req	Interviews relentlessly to break down complex requirements and resolve design dependencies
create-git-issue	Synthesizes a PRD from context, then creates dependency-aware tracer-bullet implementation issues
tdd-implementation	Enforces red-green-refactor with behavior-focused tests and thin vertical implementation slices
run-with-it	Routes execution by complexity/capability, selects agent+model from registry, and invokes the unified run-agent.sh runner

Runtime Assets

File	Purpose
assets/agent-registry.json	Agent aliases, detection commands, invocation templates, model catalog, and routing metadata.
assets/run-agent.sh	Bash runner used by macOS, Linux, Git Bash, and GUI-launched Unix-like agent workflows.
assets/run-agent.ps1	PowerShell runner for Windows workflows.
assets/prompt.md	Shared execution prompt used by the runner workflow.
assets/review-prompt.md	Review prompt material for follow-up quality gates.
assets/complexity-prompt.md	Complexity scoring prompt material used by routing workflows.

Tests

The test suite is mostly shell-based contract coverage around installer behavior, runner behavior, routing documentation, and skill workflow rules.

Focused commands:

bash tests/install-assets-contract.test.sh
bash tests/uninstall-contract.test.sh
bash tests/run-agent.test.sh
bash tests/break-req-contract.test.sh
bash tests/create-git-issue-routing.test.sh
bash tests/run-with-it-routing.test.sh

Note: tests/add-two-numbers.test.sh is a legacy tracer-bullet test file; the calculator script it references is not part of the current codebase.

---

Adding a Skill

1. Create skills/<name>/SKILL.md with YAML front-matter:

---
name: skill-name
description: What this skill does and when to use it
---

<!-- rules, examples, and any supporting content below -->

2. Add your rules, examples, or reference material in the body.
3. Commit and push — the skill is immediately available to all agents on next install/update.

---

Repo Structure

AI-Skills/
├── README.md               # Project overview, install docs, and repo map
├── LICENSE
├── install.sh              # Smart one-liner installer
├── install.ps1             # Windows installer
├── uninstall.sh            # Full uninstaller for fresh reinstalls
├── uninstall.ps1           # Windows uninstaller
├── gemini-extension.json   # Gemini CLI extension manifest
├── technical_requirements.md
├── assets/                 # Shared prompt + runner scripts
│   ├── agent-registry.json
│   ├── complexity-prompt.md
│   ├── prompt.md
│   ├── review-prompt.md
│   ├── run-agent.ps1
│   └── run-agent.sh
├── skills/
│   ├── break-req/
│   │   └── SKILL.md
│   ├── create-git-issue/
│   │   └── SKILL.md
│   ├── run-with-it/
│   │   └── SKILL.md
│   ├── save-tokens/
│   │   └── SKILL.md
│   └── tdd-implementation/
│       └── SKILL.md
└── tests/
    ├── add-two-numbers.test.sh       # Legacy tracer-bullet test
    ├── break-req-contract.test.sh
    ├── create-git-issue-routing.test.sh
    ├── install-assets-contract.test.sh
    ├── run-agent.test.sh
    ├── run-with-it-routing.test.sh
    └── uninstall-contract.test.sh