triage-cleanup

Janitor for the auto-triage pipeline. Scans tickets left CLAIMED (carrying the auto-triage claimMarker) whose claim is older than a threshold — a run that crashed, timed out, or was abandoned — and reconciles each to its true state: a ticket that already has an open auto-PR is healthy (→ IN_REVIEW, or left CLAIMED — never reaped), one with an attached plan issue goes to NEEDS_HUMAN, and only a genuinely orphaned one (no PR, no plan) returns to the queue for retry. Use as `triage-cleanup [<minutes>]`, or when asked to unstick / reap stale auto-triage claims. Tracker- and project-agnostic; reuses the triage config + tracker binding. Read-mostly: only posts a comment and a status transition — never edits code, never opens PRs.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

67%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-structured, actionable maintenance procedure with precise routing logic, explicit output contract, and conservative guardrails. It is slightly redundant between the intro, procedure, and guardrails, and leans on the sibling triage skill's reference files rather than carrying its own bundle.

Suggestions

Trim the two intro paragraphs and the Guardrails section to remove points already covered in the Procedure (e.g. the 'open auto-PR means healthy — never reap' rule appears twice), improving conciseness.

Add an explicit post-transition verification step (e.g. re-read the item to confirm the new state) to turn the implicit reconcile-then-act pattern into a visible validation checkpoint for this batch operation.

Either vendor the tracker-binding and work-item-schema references into this skill's own references/ directory, or state the cross-skill dependency explicitly at the top so the navigation path is unambiguous.

Dimension	Reasoning	Score
Conciseness	Mostly efficient and assumes Claude's competence (no explaining what a PR or tracker is), but the two motivational intro paragraphs and the Guardrails section restate points already made in the Procedure, which could be trimmed.	3 / 5
Actionability	Concrete and executable for an instruction-only skill: names exact config files, states (CLAIMED/IN_REVIEW/NEEDS_HUMAN/QUEUED), capability calls (setState/comment/hasOpenAutoPR), and the literal JSON output shape; minor gap is that actual per-tracker command syntax is delegated to tracker-binding.md.	4 / 5
Workflow Clarity	Clear three-step sequence (find candidates → reconcile via ordered sub-cases → emit JSON) with explicit per-candidate verification ('read what it actually produced') and uncertainty gates ('if you can't determine claim age... leave it'); the only gap is no explicit post-transition confirmation, but idempotency and the read-mostly nature mitigate this.	4 / 5
Progressive Disclosure	Well-organized into Invocation/Procedure/Guardrails with a 'Read first' block pointing one level deep to config and capability-contract references that are clearly signaled; references live in the sibling `triage` skill's bundle rather than this skill's own, a minor organizational gap.	4 / 5
	Total	15 / 20 Passed

Description

88%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, specific description that states what the skill does, when to use it, and its scope/side-effect boundaries. Trigger phrasing is explicit and natural; the only gap is slightly broader synonym coverage and the inherent proximity to the `triage` skill.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions with precise routing logic — 'reconciles each to its true state', open auto-PR '→ IN_REVIEW', plan issue '→ NEEDS_HUMAN', orphaned 'returns to the queue', and 'only posts a comment and a status transition — never edits code, never opens PRs'.	5 / 5
Completeness	Explicitly answers both 'what' (scans stale CLAIMED tickets and reconciles each to its true state with named outcomes) and 'when' ('Use as `triage-cleanup [<minutes>]`, or when asked to unstick / reap stale auto-triage claims').	5 / 5
Trigger Term Quality	Gives explicit triggers — 'when asked to unstick / reap stale auto-triage claims' — with natural verbs (unstick, reap, stale), but coverage is good rather than comprehensive; a few common phrasings (e.g. 'stuck claims', 'abandoned tickets') are absent.	4 / 5
Distinctiveness Conflict Risk	Carves a clear niche ('Janitor for the auto-triage pipeline', cleanup vs. processing) with distinct triggers, but shares config and tracker semantics with the sibling `triage` skill, leaving minor overlap risk.	4 / 5
	Total	18 / 20 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: MrToxy/claude-skills
Path: triage-cleanup/SKILL.md
Commit: 66c7102

Reviewed: 6 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.