luisfarfan
UserKeep your Claude Code agents lean and in control — bounded agent loops, per-agent model routing, and real per-model token measurement.
Categories
Indexed Skills (6)
loop
Execute a planned change with the Rein loop: bounded fresh-agent implementation steps, then an independent review gate, driven by this project's flow.config.json. Use when the user asks to implement, execute or run an already-planned change, or invokes /rein:loop.
plan
Plan a change into verifiable tasks with acceptance criteria, dependencies and a bounded verification command each — dry-run and explicit confirmation before anything is written. Never implements. Use when the user asks to plan or break down work, or invokes /rein:plan.
review
Independent review of a completed change — the third role, distinct from planner and implementer. Runs the mechanical checks, delegates the five-axis judgement to the code-reviewer, and records a state-bound verdict. Never implements. Use when the user asks to review a change or audit a branch before merge, or invokes /rein:review.
role
Assign this session's working role — planner, implementer or reviewer — and load its operating profile. Use when the user runs /rein:role, says which role a session should play, or when you are about to plan, implement or review and it is unclear which one you are.
run-auto
Repeat /rein:run's per-task procedure in a bounded loop until a deterministic stop condition — nothing ready, a verification failure, a human-review gate, or the per-pass cap. Use when the user asks to work through several tasks in one go, or invokes /rein:run-auto.
run
Execute exactly ONE ready task of the plan, verify it, and only then tick its checkbox. One task per invocation, with hard attempt limits. Use when the user asks to implement or run the next task, or invokes /rein:run.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.