llms.txt@docs git:20261014.72a9176A

Immutable. This exact content never changes and is served at /api/v1/blob/63d5b568ad4de892.

# agent-playbooks

> Portable, tool-agnostic instructions for AI coding agents, built on the AGENTS.md standard. One instruction set works across Claude Code, Cursor, Antigravity, Codex CLI, GitHub Copilot, Gemini CLI, Devin CLI, Factory Droid, Grok Build, Kimi Code, OpenCode, Pi, and Hermes Agent -- no per-tool configuration needed.

agent-playbooks solves two problems for teams using AI coding agents: every tool wants instructions in its own format, and an agent left to its own judgment fails in predictable ways (guessing instead of asking, grading its own work, forgetting conventions between sessions, running destructive commands). It provides 41 playbooks -- reproduce-before-fix bug fixing, test-first feature development, independent code review, security review, frontend/backend testing, cold-start exploratory QA, safe database migration, incident response, and autonomous operation -- each verified against a real test case before publishing, installed with a single shell command and no account or token.

## Docs

- [README](https://github.com/chiragmangaldev3112/agent-playbooks#readme): full install instructions, per-tool wiring notes, and the complete playbook index
- [CHANGELOG](https://github.com/chiragmangaldev3112/agent-playbooks/blob/main/CHANGELOG.md): every released version and what changed
- [THIRD_PARTY](https://github.com/chiragmangaldev3112/agent-playbooks/blob/main/THIRD_PARTY.md): every tool and hosted service the playbooks/scripts actually invoke, with licences and credited projects
- [Teaser video](https://chiragmangaldev3112.github.io/agent-playbooks/teaser.html): a ~95-second tour of the engineering-loop discipline -- one AGENTS.md, classify-route-verify, bug-fix, feature-development, code review, independent verification, the safety guardrail, blind build/verify, and codebase mapping
- [Full narrated demo](https://chiragmangaldev3112.github.io/agent-playbooks/demo.html): 36 of the 41 playbooks narrated (all but config-protection, finish-work, blind-build-verify, requirements-store and test-case-tracker, added after recording), each with a real terminal run
- [SECURITY](https://github.com/chiragmangaldev3112/agent-playbooks/blob/main/SECURITY.md): the actual trust model -- every release is signed and its file hashes verified before install, what the safety guardrail does and doesn't cover, how to report a vulnerability
- [CONTRIBUTING](https://github.com/chiragmangaldev3112/agent-playbooks/blob/main/CONTRIBUTING.md): what's in this repo to contribute to and how to test a change before opening a PR

## Install

curl -fsSL https://raw.githubusercontent.com/chiragmangaldev3112/agent-playbooks/main/install.sh -o install.sh && chmod +x install.sh && ./install.sh /path/to/your/project

On Windows run the commands in Git Bash or WSL, not PowerShell (curl and chmod differ there), and install jq first (winget install jqlang.jq). Details: https://github.com/chiragmangaldev3112/agent-playbooks#windows