Content
71%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 well-structured, largely self-contained analytical workflow with concrete output formats, exact taxonomies, and a clear mode-selection table. Its one real defect is that two of the five modes depend on reference files that do not exist in the bundle, breaking navigation for exactly the modes the description promotes.
Suggestions
Create the missing references/knowledge-map.md and references/feynman.md files (or inline their essential procedures into SKILL.md) so the map and feynman modes are actually executable as written.
Trim the Anti-Patterns section of items that restate workflow rules already stated above (e.g., data provenance, prompt injection) to reduce token cost without losing information.
Inline a minimal fallback for each referenced mode — e.g., a one-line description of the map node/edge types or the Feynman recall cycle — so the skill degrades gracefully when the reference files are unavailable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is instruction-dense — mode table, evidence-ledger field table, exact confidence labels, a good/bad central-claim example — with almost no explanation of concepts Claude already knows. Minor redundancy: the Anti-Patterns section restates rules already given in the workflow (data provenance from step 5, prompt injection from the header). | 4 / 5 |
Actionability | Concrete, executable guidance throughout for an instruction-only skill: exact ledger fields, exact relationship types (supports, explains, qualifies, contradicts, illustrates), exact four confidence labels, and a worked good/bad example. The gap is that the map and Feynman deliverables defer to reference files that are not present in the bundle. | 4 / 5 |
Workflow Clarity | Seven steps are clearly sequenced with an explicit up-front validation checkpoint ("Confirm that extraction is usable before analyzing it" plus a stop condition if extraction is incomplete). Not a destructive or batch operation, so the validation cap does not apply, but there is no fix-and-retry feedback loop beyond the source check. | 4 / 5 |
Progressive Disclosure | Section structure is clear and references are well signaled in the mode table and workflow, but the bundle contains no references/ directory — both references/knowledge-map.md and references/feynman.md are dangling, so the map and feynman modes cannot actually follow their prescribed procedures. | 3 / 5 |
Total | 15 / 20 Passed |