Content
85%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 operational skill: concrete snouty commands, gated preflight validation, branching workflows with feedback loops, and exemplary progressive disclosure into a verified one-level reference bundle. Its only weaknesses are mild verbosity in the follow-up-skills section and a handful of inherently open-ended analysis steps.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient and assumes competence — no explanations of concepts Claude already knows, and dense operational prose throughout. Minor trimming is possible: the five-skill "Suggesting follow-up skills" section (including the long mutation-testing gating conditions) and the duplicated version/compatibility details push it past lean, matching anchor 4's "minor instances of over-explanation that could be trimmed" rather than anchor 5's every-token-earns-its-place. | 4 / 5 |
Actionability | Concrete, executable commands appear throughout ("snouty doctor --json", "snouty runs --json properties", "snouty runs --json show ${RUN_ID}", "snouty runs --json build-logs ${RUN_ID}", "snouty runs show <RUN_ID> --web") plus a jq pipeline example and explicit decision rules for edge cases. A few steps remain open-ended ("Analyze the downloaded log locally", "Deeply investigate the failure") with execution detail deferred to reference files, so it is mostly-executable anchor 4 rather than fully copy-paste-ready anchor 5. | 4 / 5 |
Workflow Clarity | Each workflow is a numbered sequence with explicit validation checkpoints and error recovery: the preflight gates ("Proceed only when the top-level `ok` is `true`"), the DO-NOT-PROCEED prerequisites, conditional branches (triage_report null → fetch properties anyway; 404 → report and stop), the retry loop of downloading more counterexample and passing logs when uncertain, and a closing Self-Review checklist. This matches anchor 5's explicit validation, feedback loops, and checklists; the workflows are read-only triage so the destructive/batch cap does not apply. | 5 / 5 |
Progressive Disclosure | SKILL.md is a clear overview that explicitly instructs "Do NOT read them all up front — only read a reference file when you are told to" and names each of the six reference files at the point of need; all six exist in references/ (verified) and are one level deep with bulk detail (e.g., the 17KB logs guide) appropriately split out. The only cross-links are two clearly signaled sibling pointers inside properties.md, which do not bury content, so this matches anchor 5 rather than the minor-gaps anchor 4. | 5 / 5 |
Total | 18 / 20 Passed |