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

agent-triage

Skill10 starsChanged 40 days agoMITView on GitHub

Use when the user says "triage", "이슈 정리", "분류", or wants to scan open llmux GitHub issues, decide which are safe for an agent to implement, label them, and open a tracking branch + draft PR for each ready one. Reads and labels issues; never writes source code.

  • Commits and pushes

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. HermeticOrmus/ormus-agent-ops★ 0same name
    2. Ormus-Solutions/ormus-agent-ops★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository7 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

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

    Leave the worktree for agent-resolve to reuse, or git worktree remove it — the pushed branch is what the draft PR needs.

    line 43

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “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-triage — gate issues for the agent pipeline

agent-triage is a default-reject safety gate, not a classifier. An issue earns ready-to-agent only when all 7 gate items have citable evidence. Everything else gets an explicit label — never silently skipped. For each ready issue it opens a draft PR that acts as the work lock for agent-resolve. Pairs with agent-resolve (does the work) and agent-loop (orchestrates both).

Repo: 2lab-ai/llmux, base branch main. Labels were created already (ready-to-agent / needs-human / needs-design / not-in-repo / agent-wip / agent-blocked).

Steps

  1. Fetch un-triaged open issues. gh issue list --repo 2lab-ai/llmux --state open --json number,title,labels,body. Skip any issue that already has one of the six agent labels — labels are the pipeline's memory; re-triaging them is thrashing (a human removes the label to re-enter).
  2. Apply the 7-point gate to each, citing evidence per item. ALL must pass for ready-to-agent:
    1. In-repo completeness — name ≥1 candidate file:path inside llmux that holds the root cause. Can't name one → fail.
    2. Spec clarity — a one-sentence done-condition exists.
    3. No human/product/policy decision required.
    4. No security surface — no secret/auth/credential/cost/rate-limit changes.
    5. No release surface — no Cargo.toml version, v* tags, brew formulae, or .github/workflows/.
    6. Verifiable by just check (fmt + clippy -D warnings + tests), possibly with a new in-repo test.
    7. Bounded blast radius — roughly one module; not a broad re-architecture.
  3. Pick the label from the first failure:
    • all 7 pass → ready-to-agent
    • fails #1 (root cause outside llmux) → not-in-repo
    • fails #2 or #7 (ambiguous / too large) → needs-design
    • fails #3 / #4 / #5 (needs a human, security, or release call) → needs-human
  4. Post the gate evidence as an issue comment (a 7-row pass/fail table with the cited evidence) — this is the auditable artifact — then apply the label. The comment must exist before ready-to-agent.
  5. For each ready-to-agent issue, open the work lock (idempotently) — never in the main checkout:
    • Idempotency: if git ls-remote --heads origin "agent/issue-<N>-*" returns a branch, skip (already tracked).
    • Else create the tracking branch in a throwaway worktree so the main tree is untouched (same path convention agent-resolve reuses — <repo>/../llmux-wt/issue-<N>): git -C <repo> worktree add -B agent/issue-<N>-<slug> "<repo>/../llmux-wt/issue-<N>" main → git -C "<wt>" commit --allow-empty -m "chore(agent): track #<N> <slug>" → git -C "<wt>" push -u origin agent/issue-<N>-<slug>. Leave the worktree for agent-resolve to reuse, or git worktree remove it — the pushed branch is what the draft PR needs.
    • gh pr create --draft --base main --head agent/issue-<N>-<slug> --title "<type>: <issue title>" --body "Refs #<N>" (draft PR linking the issue; the draft state is the "not yet worked" signal agent-resolve keys on).
  6. Report a table: every open issue → assigned label → (branch + draft PR link if ready). Surface borderline issues explicitly (the needs-* / not-in-repo ones), don't bury them.

Red flags — STOP

  • Marking ready-to-agent without a cited file:path — that is trusting your own judgment over evidence. Default is reject.
  • Silently skipping a hard issue instead of labeling it needs-design / needs-human / not-in-repo.
  • Re-triaging an already-labeled issue (thrashing).
  • Creating a second branch/PR for an issue that already has one (always ls-remote first).
  • Editing any source file — agent-triage reads and labels only.

Example: why issue #1 is NOT ready

The "gpt-5.5 remaining-context" issue fails gate item #1: its root cause is in the Claude Code client (it infers the context window from the model-name string); no llmux file:path can fix the displayed number. → label not-in-repo, comment with the evidence, no PR. Shipping a "fix" here would be a non-fix. This is the exact failure mode the gate exists to catch.

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.