- five-minute-maintainer-loop.md 45 lines · 2.4 KB
Five-Minute Maintainer Loop
Loop Engineering — the
/loopverb repeats a task while you are present. This one does continuous, low-risk upkeep on a tight timer while you focus on the real work. Inspired by Peter Steinberger's five-minute maintainer.
🎯 Goal
Every few minutes, make exactly one small, verified improvement to the repo — and let the agent decide what to clean. That judgment is the whole point; nothing is hardcoded.
⏱️ Schedule
Suggested interval: 5m (runs while your session is open).
▶️ Run it
/loop 5m "Make one small, verified repository improvement: a flaky test, a stale comment, a missing type, a dead import. One change, one commit, tests green. Never touch anything risky or architectural. If nothing safe remains, say so and wait."🔁 Iteration steps
- Perceive — scan for a small, low-risk improvement opportunity.
- Reason — confirm it is safe; the
scope-guardhook blocks risky/out-of-scope edits. - Act — make the single change.
- Verify — run the suite with
test-runner; only proceed if green. - Observe — commit with
/commit; one change, one commit. Wait for the next tick.
🛑 Stopping condition
You end the session, or there is no safe improvement left this pass (the loop idles rather than inventing risky work).
💰 Budget & guardrails
One change per tick; never refactors or touches production-critical paths. Set a per-session token budget before walking away.
🧩 Referenced components
agent:development-team/test-runner— verifies each change is green.command:git-workflow/commit— one clean commit per improvement.hook:quality-gates/scope-guard— blocks risky / out-of-scope edits.
💡 Example
While you build a feature, the loop quietly fixes a flaky timer test, removes a dead import, and adds a missing return type — three tiny green commits — without ever touching the feature you're working on.
Scanned by NVIDIA SkillSpector
Automated static analysis (no LLM) · v2.12.0 · 2026-10-17
SkillSpector looks for patterns such as shell commands, credential access, network calls or unpinned dependencies. An observation is something worth reading in the files before you install, not a confirmed vulnerability. Components that run code locally, like hooks, MCP servers and security tooling, naturally show more of them.
No observations.
Learn more about NVIDIA SkillSpector or how components are reviewed.