| status | draft |
|---|---|
| audience | individual |
Personal workflow: Plan → Ship → Spec. How AI dev tooling integrates into your daily loop.
┌─────────────┐ ┌──────────┐ ┌──────────┐ ┌────────┐
│ Plan │──────▶│ Ship │──────▶│ Spec │──────▶│ Review │
│ (CLAUDE.md) │ │ (Skills) │ │(Handoff) │ │(Tools) │
└─────────────┘ └──────────┘ └──────────┘ └────────┘
▲ │
└───────────────────────────────────────────────────────────┘
Copy a rule file into your project:
rules/CLAUDE.md→ Claude Code, Codexrules/COPILOT.md→ GitHub Copilotrules/GEMINI.md→ Gemini CLIrules/AGENTS.mdfor multi-agent flows
File is always loaded. You only write once, reuse forever. See Conventions as Code for how rules layer on tool defaults.
Skills are verb-named, single-purpose tools. Activate them in your prompt:
/recall "feature name"— find relevant code patterns/context-pack "module path"— assemble API context/plan "5-step refactor"— break down large changes/dispatch "2 subagents"— parallelize analysis
See Kit Overview for the full skill catalog. Link a skill in your prompt; the AI tool calls it automatically.
When shipping to production, commit a spec:
echo "### New Feature X
- What: Async timeout for connections
- Why: Prevent resource leaks
- How: Thread timeout param through pool, add circuit breaker
" > docs/specs/2026-04-15-feature-x/spec.mdSpecs feed the auto-generated docs/roadmap.md. Over time, they become your project's decision audit trail. See Benchmarks for why this matters.
As you scale:
- Enable hooks:
bash install-rag.sh --with-hooks(optional) - Check the governance checklist: Governance
- Share For Teams with your lead if crossing into team work
Questions? See Primitives for "should this be a skill, agent, or hook?" decision tree.