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 strongly actionable, well-structured configuration skill: concrete diagnostic commands, explicit config-trust rules, and exemplary progressive disclosure into per-platform reference files. The main improvement opportunities are showing an actual `memsearch config set` example and cutting the repeated disclaimers and "see your platform reference file" reminders.
Suggestions
Add one copy-paste `memsearch config set` example showing key, value, and global vs project scope (e.g. enabling plugin summarize globally) since applying changes is the skill's central flow.
State the "MemSearch, not host memory" clarification once (as the body itself instructs) and consolidate the repeated "see your platform reference file" pointers into a single convention note.
Present the diagnostic sequence as a numbered checklist with an explicit fix-and-retry loop when `.index-state.json` reports `degraded` or `error`.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is command-first and dense with no explanations of concepts Claude already knows, but it repeats the "this is MemSearch, not the host agent's memory/config system" disclaimer three times and the phrase "see your platform reference file" roughly ten times — trimmable padding that keeps it below anchor 5, while it is clearly more efficient than anchor 3's unnecessary explanations. | 4 / 5 |
Actionability | Diagnosis is fully executable (copy-paste `memsearch config list --resolved/--global/--project`, version checks, PyPI curl, file and index-state inspection) with concrete TOML provider blocks, but the core change flow only says "Use `memsearch config set` for changes" without a single example invocation showing key, value, and scope flags, which is a minor gap versus anchor 5. | 4 / 5 |
Workflow Clarity | The flow is well sequenced — identify platform and read its reference file, route intent, run the diagnose-first command sequence, apply changes, then "show the command, the resolved value, and whether a new session is needed" — with error-state inspection of `.index-state.json` and `.maintenance-state.json` providing feedback loops and ask-first guardrails for destructive/paid operations. It falls short of anchor 5 because the diagnostic sequence is not presented as an explicit ordered checklist with fix-and-retry loops. | 4 / 5 |
Progressive Disclosure | The SKILL.md body is a well-sectioned overview that correctly pushes platform-specific detail (version-check commands, plugin keys, restart guidance) into five real, one-level-deep reference files, instructs the agent to read only the one matching its environment, and clearly signals each path; the referenced files exist and contain no nested references. | 5 / 5 |
Total | 17 / 20 Passed |