Content
82%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.
The skill delivers concrete, executable MCP guidance with a strong error-analysis framework and report template. Its main weaknesses are mild redundancy between the preferred diagnostic path and the manual fallback steps, and implicit rather than explicit validation checkpoints.
Suggestions
Consolidate Steps 2–6 into a compact 'Manual fallbacks (when harness_diagnose is unavailable)' section, noting explicitly when to prefer them over the one-call diagnose path, to trim redundancy.
Add an explicit validation checkpoint before the Response Format step, e.g. 'Confirm all failed steps were checked and chained child pipelines were followed before writing the report', to make the diagnostic loop concrete.
Move the Analysis Framework error catalog (Build/Infra/Config/Deployment/Timeout/Artifact errors) into references/error-catalog.md and link to it, keeping SKILL.md as a lean overview with the diagnose flow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient — concrete MCP call blocks, a parameter table, and error→fix mappings with no tutorial padding — but Steps 2–6 partially duplicate what 'harness_diagnose' already returns, and Performance Notes restate parameter behavior. Matches 'efficient; minor instances of over-explanation that could be trimmed' rather than the lean 5 anchor. | 4 / 5 |
Actionability | Gives copy-paste-ready MCP invocations with exact parameter names and defaults ('include_logs: true', 'log_snippet_lines: 120'), a full parameter reference table, and an Examples section mapping user requests to specific tools. Fully executable guidance covering the common cases. | 5 / 5 |
Workflow Clarity | The sequence is clear (Step 1 preferred, Step 1b deep mode, manual fallbacks) with a Troubleshooting recovery section, but checkpoints are implicit — there is no explicit step telling Claude to confirm all failed steps were examined or that chained-pipeline drill-down completed before writing the report. Fits 'clear sequence with most checkpoints present; minor validation gaps'. | 4 / 5 |
Progressive Disclosure | Well-organized sections with easy header navigation and no nested references, but at ~172 lines the Analysis Framework error catalog and manual fallback steps could be split into a reference file. Fits 'good structure; most content appropriately placed; minor organization gaps' rather than the exemplary one-level-deep-reference structure. | 4 / 5 |
Total | 17 / 20 Passed |