Content
78%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 tight, well-structured validation skill that keeps its executable weight in the bundled run-tests.sh while using the body for check definitions, pass criteria, and scope boundaries. Its main weaknesses are the pseudocode placeholder in the compilation check and a slightly redundant output-format template.
Suggestions
Replace the placeholder comments in 'Workflow Example Compilation' with the actual extraction/compile commands (or point to the exact phase in scripts/run-tests.sh that performs them) so check 2 is as executable as check 1.
Trim the 'Output Format' markdown template to the key fields, or note that the script emits this format automatically, since the full template duplicates script output.
State explicit pass/fail criteria for checks 3–5 (permission, structural, cross-workflow) the way check 2 does, so every check has a defined verification outcome.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient — sections like 'When to Use', 'Checks', and 'What This Skill Does NOT Do' each earn their tokens with no concept explanations Claude already knows. Minor trim opportunities remain: the ~20-line 'Output Format' template duplicates what the bundled script already emits, and check 2 uses placeholder comments rather than trimmed or omitted prose. | 4 / 5 |
Actionability | Mostly executable guidance: check 1 gives a copy-paste bash loop, the 'Running' section gives the exact command `bash skills/skill-tester-ci/scripts/run-tests.sh` (which exists in the bundle), and checks 3–5 state concrete greppable rules. The gap is check 2, where workflow extraction is described only as comments ('# Extract markdown code blocks with frontmatter ... # Run: gh aw compile') rather than runnable code. | 4 / 5 |
Workflow Clarity | The five checks are numbered and sequenced, matching the script's phase order, with explicit pass criteria ('Zero compilation errors. Warnings are reported but don't fail.') and a defined per-skill result report including error reporting. Validation is inherent to this report-only skill (it explicitly 'Does not modify skills — reports findings only'), so the batch-operation cap doesn't apply; the minor gap is no fix-and-retry guidance and no explicit criteria for checks 3–5. | 4 / 5 |
Progressive Disclosure | SKILL.md is a clean overview of the five checks with well-organized headers, a rules table, and a single one-level-deep bundle reference (`scripts/run-tests.sh`) that is explicitly linked in the 'Running' section and exists in the bundle. Nothing that belongs in a separate file is inlined; navigation is easy. | 5 / 5 |
Total | 17 / 20 Passed |