agentleFS
2lab-ai2lab-ai/llmux/.claude/skills/agent-loop/SKILL.md

agent-loop

Skill10 starsChanged 40 days agoMITView on GitHub

Use when the user says "loop", "반복", "쭉 돌려줘", or wants to run triage then resolve repeatedly until there are no agent-workable llmux issues/PRs left. Thin orchestrator over the agent-triage and agent-resolve skills; dry-run by default.

10 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Mode: Default is DRY-RUN: run agent-triage's evaluation and print the resolve plan, change nothing.

    line 15
  2. Very specific:

    Never re-triage an issue that already carries a needs-* / not-in-repo / agent-blocked label — a human must clear the label for it to …

    line 33

“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

agent-loop — orchestrate triage → resolve until drained

agent-loop has no judgment of its own. It runs agent-triage once, then agent-resolve on the resulting ready PRs one at a time, and stops on clear conditions. Labels are its memory — it never re-evaluates issues already marked needs-* / not-in-repo / agent-blocked. Inherits agent-resolve's boundary: it does not merge or 배포.

Steps

  1. Mode. Default is DRY-RUN: run agent-triage's evaluation and print the resolve plan, change nothing. Act only if the user said "apply" / "실제로 해" / passed --apply.
  2. Run agent-triage once. (Labels issues, opens draft PRs for ready ones, skips already-labeled.)
  3. Resolve. Over ready-to-agent draft PRs, oldest first, run agent-resolve — each in its own git worktree (per that skill), so resolves are isolated and may run in parallel. Verify each PR genuinely green (just check + CI) independently before counting it done. Parallelize only across issues that touch disjoint modules; if their diffs would overlap, serialize them to avoid merge conflicts.
  4. Stop when ANY:
    • no ready-to-agent draft PRs remain (the user's "no issues/PRs left" condition),
    • --max-iterations reached (default 3),
    • 2 consecutive agent-resolve runs end in agent-blocked,
    • an agent-resolve run hit a kill-switch needing a human (needs-design / needs-human / security / release).
  5. Report a run summary: issues triaged (+ labels), PRs made review-ready, PRs blocked (+ why), and exactly what is left for a human (review → merge → deploy).

Anti-thrashing

Never re-triage an issue that already carries a needs-* / not-in-repo / agent-blocked label — a human must clear the label for it to re-enter the pipeline. Without this, the same un-fixable issue (e.g. the gpt-5.5 client-limitation one) gets re-evaluated to the same dead end every iteration.

Red flags — STOP

  • Running in apply mode without explicit user opt-in.
  • Re-triaging labeled issues (thrashing) — labels are memory.
  • Merging or deploying — agent-loop inherits agent-resolve's human-gated boundary.
  • Running unbounded — always honor --max-iterations and the stop conditions.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.