Content
62%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill body is well-sequenced with explicit verification and recovery loops, but it leans on abstract meta-framework language and thin executable examples, and its many file references are not backed by any real bundle. Tightening redundancy and making references concrete would raise the weaker dimensions.
Suggestions
Replace abstract directives ("Infer root cause", "Isolate root cause") with concrete executable steps or example commands, and expand the canonical workflow beyond two `rg` lines to a real reproduce→fix→verify command sequence.
Resolve the bundle: either create the referenced resources/*.md and ../_shared/core/*.md files and link them as one-level-deep clickable references, or remove references to files that do not exist.
De-duplicate the References section (execution-protocol.md, examples.md, and checklist.md are each listed twice) and trim meta-framework scaffolding like "SSL primitive" and "Intent signature" that does not aid execution.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly terse tables and bullets, but repeats the same reference paths (e.g. "resources/execution-protocol.md" appears in both prose and the bulleted list) and carries meta-framework jargon ("SSL primitive", "Intent signature", "Scenes") that does not earn its tokens; it could be tightened rather than being fully lean. | 2 / 3 |
Actionability | It offers some concrete bits (the `rg` canonical path and Serena `find_symbol("functionName")` calls) but most guidance is abstract — "Infer root cause", "Isolate root cause and reject unsupported hypotheses" describe rather than instruct, and the canonical workflow is only two search commands, so it falls short of copy-paste-ready executable guidance. | 2 / 3 |
Workflow Clarity | The PREPARE→ACQUIRE→REASON→ACT→VERIFY→FINALIZE sequence is explicit, with a VERIFY validation checkpoint and a feedback loop ("If the first fix fails verification, return to root-cause analysis"), matching the score-3 anchor of a clear sequence with validation and error-recovery loops. | 3 / 3 |
Progressive Disclosure | No bundle files exist under references/, scripts/, or assets/, yet the body points to many resources/*.md and ../_shared/core/*.md paths that are not real files and are not formatted as one-level-deep links; the SKILL.md itself is a ~145-line monolithic doc rather than a lean overview, so it has structure but poorly signaled, unverifiable references. | 2 / 3 |
Total | 9 / 12 Passed |