Content
86%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 exemplarily lean, well-structured process skill: the body is an overview of workflow, evidence, and curation rules while all detail lives in four focused one-level-deep reference files. The only shortcomings are minor: route-selection and check-execution specifics are unspecified, and there is no explicit re-validation loop after applying fixes.
Suggestions
Add one sentence on how to choose representative routes and what tooling runs the deterministic source checks (e.g., grep patterns for token IDs).
Add an explicit post-fix validation step, such as re-running the audit or tests to confirm findings are resolved before delivery.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~45-line body is pure directive ("Separate mechanical violations from judgment. Report suggestions as suggestions unless a documented rule makes them violations.") with no concept explanation or padding; every line instructs and nothing restates what Claude already knows. Fits the lean/efficient top anchor exactly. | 5 / 5 |
Actionability | Concrete steps with real file paths, version-matching logic ("`openclaw-carapace` for new installs or the `openclaw-design-system` compatibility alias for an upgraded lock"), and exact thresholds ("Surface at most five non-error findings"). Minor gaps: "Inspect representative rendered routes at desktop and mobile sizes" leaves route selection unspecified, and how deterministic source checks are executed is not stated in the body. Not 3 (guidance is concrete and specific, not high-level hints); not 5 (route selection and check tooling left to judgment). | 4 / 5 |
Workflow Clarity | Ten clearly ordered steps with correct sequencing ("Run deterministic source checks before judgment-based review", theme and viewport checks before report emission), and validation guardrails exist in referenced policies (fix-policy: "Preserve tests and include real browser evidence"; delivery: `--force-with-lease`, branch-ownership checks, decision table). The body itself lacks an explicit post-fix re-validation feedback loop, keeping it below 5; checkpoints are mostly present, so above 3. | 4 / 5 |
Progressive Disclosure | A lean overview body with four well-signaled, one-level-deep references (rubric.md, report-format.md, fix-policy.md, github-pr-delivery.md), all verified to exist as real files, each owning a single concern (detection rules, report schema, fix constraints, delivery procedure). No nested references beyond a sibling mention of fix-policy.md; matches the top anchor. | 5 / 5 |
Total | 18 / 20 Passed |