Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-structured, highly actionable rubric body with concrete failure signatures, rewrite tables, a clear sequenced procedure, and a properly split-out worked-examples reference. Its only weakness is verbosity: several ISTQB glossary citations and basic testing-concept explanations restate knowledge Claude already has.
Suggestions
Trim or remove the inline ISTQB glossary definitions and citation URLs for testability, expected result, test oracle, equivalence partitioning, and boundary value analysis — Claude already knows these concepts, and the heuristics stand on their own without them.
Condense the 'What testability means here' opening, which quotes ISTQB at length before the practical content begins; a one-line framing would preserve the shift-left motivation with fewer tokens.
If retaining any external definitions, move them to references/ alongside the worked examples rather than quoting them inline in the main body.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The core content (heuristic tables, severity, verdict rule, output template, anti-patterns) earns its place, but repeated ISTQB glossary citations and explanations of equivalence partitioning and boundary value analysis are concepts Claude already knows and could be trimmed, matching the score-2 'mostly efficient but includes some unnecessary explanation' anchor rather than the lean score-3. | 2 / 3 |
Actionability | For an instruction-only skill it gives concrete, executable guidance: failure-signature word lists ('fast, clean, robust, seamless...'), untestable-to-testable rewrite tables per heuristic, a numbered 6-step procedure, and a copy-paste-ready output template with a filled example, matching the score-3 'specific examples; copy-paste ready' anchor. | 3 / 3 |
Workflow Clarity | A clear 6-step sequence (split into claims -> test each against all three heuristics -> assign severity -> write a rewrite -> compute verdict -> emit findings table) with explicit severity and BLOCK/REVIEW/OK verdict rules gives an unambiguous process; this is a review skill, not a destructive/batch operation, so the missing-validation cap does not apply. | 3 / 3 |
Progressive Disclosure | SKILL.md holds the overview (heuristics, procedure, output format) and splits the three end-to-end worked examples into a clearly signaled, one-level-deep reference to the verified references/worked-examples.md, matching the score-3 'clear overview with well-signaled one-level-deep references' anchor. | 3 / 3 |
Total | 11 / 12 Passed |