Content
100%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.
An exemplary instruction-only workflow skill: terse, non-obvious procedural guidance with concrete file paths, app IDs, and validation checkpoints at every phase, plus an explicit merge gate. All four dimensions sit at the top anchors with no padding or misfiled content.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line is non-obvious policy or procedure ('Do not treat a flat comment list as proof that all review threads are resolved', 'Never mutate another app without explicit approval') with zero concept explanations Claude already knows — lean and efficient, matching the top anchor. | 5 / 5 |
Actionability | Guidance is concrete and executable despite being instruction-only: exact paths ('AGENTS.md', 'docs/openapi/latest.json'), commands ('--help' paths, '$watch-asc-pr'), a specific disposable app ID ('6759231657'), and a reproduce → failing test → smallest fix → rerun protocol. Per the rubric's instruction-only note, absence of code is not penalized when guidance is this actionable. | 5 / 5 |
Workflow Clarity | Six clearly sequenced phases (contract → isolate → verify → fix → merge gate → hand off) with explicit validation checkpoints and feedback loops: 'Reproduce each defect before changing code', 'confirm the failure... rerun the focused test', 'Re-fetch the head SHA, checks, reviews, and GraphQL threads after every push', plus a merge-gate checklist. Destructive/live operations all carry validation, so the destructive-cap rule is satisfied. | 5 / 5 |
Progressive Disclosure | No bundle files exist (no references/, scripts/, or assets/), so the ~58-line body is appropriately self-contained with well-organized sections. The external pointers it does make ('AGENTS.md', '$watch-asc-pr') are clearly signaled and one level deep, and no inline content belongs in a separate file. | 5 / 5 |
Total | 20 / 20 Passed |