Content
75%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A lean, actionable instruction-only skill with a well-sequenced procedure, explicit stop/patch validation checkpoints, and clean references to bundled assets. Its main weakness is reliance on an external autopilot-comment skill and some deliberate-but-trimmable rule repetition.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Terse, imperative prose ("Missing field -> stop", "One PR. Do not nest a scheduler path.") assumes Claude's competence and earns its tokens, though rules like the forbidden status labels and "never assign, never request reviewers" recur across sections and could be trimmed. | 4 / 5 |
Actionability | Concrete guidance throughout — exact labels, exact template fields, the probe path "$REPO_ROOT/.agents/skills/autopilot-comment/SKILL.md", and explicit call prohibitions ("Do not call gh pr comment yourself") — but execution depends on the external autopilot-comment skill whose behavior is not bundled, leaving minor gaps short of copy-paste-ready 5. | 4 / 5 |
Workflow Clarity | A clear 6-step Procedure with explicit validation checkpoints ("Missing number -> stop", "Missing -> stop. Do not post") and an error-recovery loop ("failed post -- patch via autopilot-comment (intent: patch)"); a few checkpoints remain implicit, keeping it just below the fully-checklisted 5. | 4 / 5 |
Progressive Disclosure | Well-sectioned overview (Activation card, Origin and writes, Context, Procedure, Hard nos) pointing to real one-level-deep assets (label-contract.json, pr-triage-template.md) with clear navigation; minor organization gaps (some inline rule repetition) keep it short of 5. | 4 / 5 |
Total | 16 / 20 Passed |