Structured Markdown commands for AI-assisted development. Covers the SDLC from planning to production operations. Tool-adaptable.
Structured commands organized by SDLC phase, for engineers who want repeatable structure without ceremony:
- Two foundations:
/pb-preamble(how to collaborate) and/pb-design-rules(what to build) - A three-command daily ritual:
/pb-preferences --setup→/pb-start→ code →/pb-review→/pb-pr - Self-evolving: periodic cycles plus capability-triggered updates when the underlying model tier changes
Content is Markdown. Structure is enforced by lint and tests. Commands are versioned independently; no breaking changes without a major release.
- Engineers who want a curated command library for AI-assisted development
- Teams adapting playbooks to different AI tools (read as Markdown, adapt prompts)
- Solo maintainers building personal workflows they can evolve
Not a framework to adopt wholesale. Start with the preamble, pick commands that fit, ignore the rest.
git clone https://github.com/vnykmshr/playbook.git
cd playbook
./scripts/install.shinstall.sh symlinks commands/ and helper scripts to ~/.claude/. Using a different toolchain? Read the files directly -- each command is self-contained Markdown.
Uninstall: ./scripts/uninstall.sh.
/pb-preferences --setup # one-time, 15 min
/pb-start "<description>" # create branch, lock scope
[code]
/pb-review # self-review + persona review + commit
/pb-pr # create PR when peer review needed
For deeper work: /pb-plan (architecture), /pb-adr (decision records), /pb-security, /pb-performance, /pb-incident. See the command index for the full list.
- Command index -- complete catalog grouped by category
- Book site -- distraction-free reading view
- Integration guide -- how commands compose
- Using with other tools -- adapting the playbook outside the reference toolchain
- CONTRIBUTING -- local checks, PR expectations
- SECURITY -- scope, reporting
- CHANGELOG -- release history
- MIT -- see LICENSE