Content
67%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, actionable skill body with a clear sequenced workflow, executable commands, and a properly signaled one-level-deep reference bundle. Its main weakness is token efficiency: Principles, the report format, and the configurable-path notes are duplicated between SKILL.md and reference.md.
Suggestions
Remove the 'Principles' section (or reduce it to a one-line pointer) — it duplicates the 'Key Principles' section of references/reference.md nearly verbatim and costs context on every load.
Replace the inline 'Report Section Format' skeleton with a pointer to the report template in references/reference.md, keeping only the output filename convention inline.
State the project-configurable output/frameworks path once (e.g. in a single 'Configuration' note) instead of repeating the caveat across Principles, Steps, and Gotchas.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient, but the 'Principles' section duplicates the reference file's 'Key Principles' nearly verbatim, 'Report Section Format' repeats the reference's report template in skeleton form, and the project-configurable-path note is stated three times. It is above anchor 2 (no pervasive padding) but below anchor 4 because this duplication is more than a minor trim. | 3 / 5 |
Actionability | Concrete executable guidance throughout: 'bash ${CLAUDE_PLUGIN_ROOT}/scripts/retrospect-load-sessions.sh $@' with a parse contract ('First line: PERIOD: YYYY-MM-DD_to_YYYY-MM-DD'), a concrete output path example, and four ready-to-run usage examples. Minor gaps (e.g. the zero-sessions guard sits in 'When Not to Use' rather than as a checkpoint in the workflow) keep it below fully copy-paste-complete anchor 5. | 4 / 5 |
Workflow Clarity | Seven clearly sequenced steps with an explicit fallback checkpoint in step 1 ('If missing: warn, use generic framework') and a gotcha inviting direct testing of the loader ('Run the session loader directly to test it'). It is not a destructive or batch-modifying operation so the cap-at-3 rule does not apply; it misses anchor 5 because there is no validation step confirming the written report (e.g. verifying the output file exists or that sessions were actually found inside the workflow). | 4 / 5 |
Progressive Disclosure | Scored against the actual bundle: references/reference.md exists and the body's '[Reference](references/reference.md) — scoring rubrics, metric definitions, and report format' is one level deep and clearly signaled; domain frameworks correctly live in the reference with step 1 pointing to them. Minor gaps — the Principles list and report skeleton duplicate reference content inline — keep it below the well-split anchor 5. | 4 / 5 |
Total | 15 / 20 Passed |