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 well-structured, actionable skill body: a numbered triage workflow with confirmation gates, executable commands, and genuine one-level-deep bundle references that all exist on disk. The main weaknesses are one incorrect Quick Command path and validation scripts that are listed but never explicitly placed in the step sequence, plus mild redundancy across the Mindset, Workflow, and Anti-Patterns sections.
Suggestions
Fix the Quick Command `cp templates/REFERENCE-tool.md references/<slug>.md` — the actual template is `assets/templates/REFERENCE-tool.yaml`, so either correct the path or replace the cp with an instruction to create the file from the template's required_fields.
Wire validation into the workflow sequence, e.g. add to step 5: 'After writing references/<slug>.md, run ./scripts/validate-reference-tool.sh references/<slug>.md and fix any errors before updating REVIEWED.md' (and the analogous validate-analysis-tool.sh call in step 10).
Trim redundancy between the Mindset section, the numbered Workflow, and the Anti-Patterns — the duplicate-check and confirmation rules are each stated three times; consolidating them would sharpen the token budget.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes competence — no concept explanations, no filler — but there is some redundancy: the duplicate-check and confirmation rules each appear in Workflow, Quick Commands, and Anti-Patterns, and the Mindset section partially restates workflow rules. Minor trimmable content, matching the 'efficient with minor over-explanation' anchor rather than the every-token-earns-its-place 5. | 4 / 5 |
Actionability | Mostly executable guidance: concrete grep/curl/git-submodule commands, real template and validator paths, and a concrete REVIEWED.md row format. Minor gaps keep it below 5: the Quick Command `cp templates/REFERENCE-tool.md references/<slug>.md` references a path that does not exist (templates live at assets/templates/ and are .yaml), and the curl snippet hardcodes the /main/ branch only. | 4 / 5 |
Workflow Clarity | A clear 10-step sequence with real checkpoints (duplicate grep with "If already present, report it and stop", confirmation gates for vendoring and promotion). However, the validator scripts are listed only in Quick Commands and the References section — no step explicitly says to run them after creating a file, so validation is available but not wired into the sequence, matching 'most checkpoints present; minor validation gaps'. | 4 / 5 |
Progressive Disclosure | SKILL.md is a well-signaled overview and every referenced path exists one level deep in the bundle: references/classification-guide.md for the taxonomy, assets/templates/*.yaml for required fields, assets/schemas/*.json, and scripts/validate-*.sh. Content is appropriately split rather than inlined, and the References section maps each artifact to its purpose. | 5 / 5 |
Total | 17 / 20 Passed |