Content
96%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.
An exceptionally dense and executable troubleshooting playbook: a doctor-first workflow with a finding→cause→action mapping table, exact commands for every probe and fix, safety rules (never print keys, ask before touching the user's server), and explicit rerun/confirm validation loops. The only structural gap is that its sole progressive-disclosure reference (reference.md) is not present in the reviewed bundle, so the link cannot be verified and detail depth is unverifiable.
Suggestions
Ship reference.md alongside SKILL.md (or correct the link target) — the body's only detail offload, "More symptoms, exact error strings and log stage names: [reference.md](reference.md)", currently resolves to no file in the skill bundle.
Include the scripts/ the workflow depends on (ov-memory-doctor.mjs) or note in SKILL.md that it ships with the plugin install, so the primary Step 1 command is verifiable from the skill itself.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Nearly every line carries non-obvious, plugin-specific knowledge Claude cannot be assumed to have ("claude plugin update is a no-op when the version string did not change", "One 4xx mentioning 'mode' pins recall for 6h", "/health answers 200 even with a bad key"). There is no padding and no explanation of concepts Claude already knows — it assumes competence and every token earns its place, matching the level-5 anchor; level 4 would require over-explanations that could be trimmed, and none are evident. | 5 / 5 |
Actionability | Everything is copy-paste executable: the doctor invocation with flags (`node ${CLAUDE_PLUGIN_ROOT}/scripts/ov-memory-doctor.mjs --json --offline`), exact curl probes with headers and payloads, exact CLI remediations (`claude plugin marketplace update openviking && claude plugin uninstall ... && claude plugin install ...`), exact file paths, and a finding→meaning→action table. This matches the level-5 anchor (fully executable commands covering the common cases) rather than level 4, since no significant execution details are missing. | 5 / 5 |
Workflow Clarity | The sequence is explicit and ordered: run the doctor, read the whole report before acting, then "Work top-down; fix the first ✗ and rerun before chasing the next", then targeted checks "only when the report is not conclusive", and finally "Confirm the fix by rerunning the doctor, then by observing an <openviking-context> block on the next prompt" — a complete feedback loop with validation checkpoints. This matches the level-5 anchor; the level-3 destructive/batch cap does not apply since the workflow is read-only and validated. | 5 / 5 |
Progressive Disclosure | Sections are well organized and the single external pointer — "More symptoms, exact error strings and log stage names: [reference.md](reference.md)" — is one level deep and clearly signaled. However, no bundle directory exists in this skill (no references/, scripts/, or assets/), so the referenced reference.md (and the ov-memory-doctor.mjs the workflow depends on) cannot be verified to exist, leaving a dangling reference; this fits the level-4 anchor (good structure, references mostly clear, minor organization gaps) rather than level 5's easy, fully-resolvable navigation. | 4 / 5 |
Total | 19 / 20 Passed |