Content
92%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 tight, highly executable skill body: every step carries a concrete command, the workflow is well-sequenced with evaluation checkpoints and error-recovery branches, and there is no padding or beginner-level explanation. The only flaw is slight redundancy in the fallback/path-resolution notes.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence (no concept explanations, direct commands), but has minor trimmable redundancy: the "If `memsearch` is not found, try `uvx memsearch` instead" fallback appears twice (Steps 1 and 4), and the paragraph "The path above is relative to this skill's directory. If the host displays the placeholder literally, resolve..." re-explains path resolution already implied by the `${CLAUDE_SKILL_DIR}` variable. This fits the score-4 anchor 'minor instances of over-explanation that could be trimmed' rather than score 5's 'every token earns its place'. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready commands with all flags specified: `memsearch search "<query>" --top-k 5 --json-output --default-collection <collection>`, `memsearch expand <chunk_hash> --default-collection ...`, `memsearch transcript <jsonl_path> --turn <uuid> --context 3`, plus the `derive-collection.sh` fallback and the `uvx memsearch` alternative. Matches the score-5 anchor 'copy-paste ready code or commands' — score 4's 'minor gaps' doesn't apply since every step has a concrete command. | 5 / 5 |
Workflow Clarity | A clear 5-step sequence (Search → Evaluate → Expand → Deep drill → Return) with an explicit selection checkpoint ("Skip chunks that are clearly irrelevant or too generic") and error-recovery feedback loops for two failure modes ("If `memsearch` is not found, try `uvx memsearch`"; "If `memsearch transcript` reports an unrecognized transcript format... read the referenced file directly"). This matches the score-5 anchor's 'explicit validation steps; feedback loops for error recovery'; the operations are read-only so no destructive/batch cap applies. | 5 / 5 |
Progressive Disclosure | The skill is a ~50-line self-contained instruction sheet with well-organized sections (Project Collection, Steps, When unsure, Output Format) and no unnecessary inline bulk — the simple-skill case where a single file scores 5. Its one external reference (`../../scripts/derive-collection.sh`) is clearly signaled as a conditional fallback with resolution guidance; no nested reference chains exist. | 5 / 5 |
Total | 19 / 20 Passed |