ChristopherHoward1
UserA plan→implement→review→release→retro loop for agentic coding: Claude Code orchestrates, isolated worktree implementers build, cross-model reviewers gate every stage
Categories
Indexed Skills (7)
2-implement
Dispatch the implementer agent into an isolated worktree for an approved work unit, then drive the gate loop until green. Use after /1-plan approval.
3-review
Run an independent code review of an implemented work unit via a fresh code-reviewer subagent, then drive the fix loop to a merge-ready state. Use after /2-implement passes the gate.
1-plan
Draft a work unit and have it reviewed by a fresh plan-reviewer subagent before it becomes real. Use when starting any non-trivial piece of work.
4-release
Run the release script for a work unit, then push, PR, and tag autonomously (full TRIP).
5-retro
Close a work unit by recording lessons and routing each one to the smallest durable project artifact.
compact
Shrink the hot context tier — regenerate ARCHI.md from the current codebase and prune PLAN.md back to one screen. Use when the hot tier drifts stale or exceeds ~300 combined lines.
init
Bootstrap this framework onto a project — scan the codebase, generate ARCHI.md, fill config.yaml, and verify the gate runs. Use once when adopting the framework, or on a freshly cloned template.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.