Content
90%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 body is an exemplary lean operator guide: copy-paste-ready commands, an explicit validate-and-stop gate, a manual verification checklist, and details correctly pushed down into the bundled script. The only notable gaps are the missing error-recovery loop around validation and a dangling templates/findings.md reference that has no corresponding file in the bundle.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~50-line body is dense and lean: every rule carries operational substance (output paths, citation and sidecar policy, API-key handling, mode semantics) and nothing explains concepts Claude already knows. Even the flag explanations (--additional-query, --include-domain) document non-obvious tool-specific behavior, so every token earns its place. | 5 / 5 |
Actionability | The Running section gives copy-paste-ready commands covering report (both inline-query and --query-file forms), json, validate, and doctor, with concrete modes and flag semantics; the Validation section repeats the exact validate invocation and describes its output contract ({ok, errors, warnings, mode, synthesis, queryCount}, exit 0). Missing-key remediation is concrete (Customize tab, .env.local, KENDEX_ENV_FILE, op:// reference). | 5 / 5 |
Workflow Clarity | The sequence is clear and gated by an explicit validation checkpoint ("Once the report and its sidecar exist, run validate and stop") plus a manual review checklist (contradicted claims, off-topic sources, unsupported recommendations). It falls short of anchor 5 because no error-recovery loop is stated — what to do when validate reports errors (fix and re-validate vs. re-run research) is left implicit. | 4 / 5 |
Progressive Disclosure | Good structure: the body stays an overview and defers flag declarations and limits to the bundled script by named symbols (modeDefaults, parseArgs, applyMode — all verified present in scripts/deep-research). It misses anchor 5 because the referenced path templates/findings.md does not exist in the bundle (the template is embedded in the script), and agents/planner.md plus ../code-quality/SKILL.md point outside the skill's own files. | 4 / 5 |
Total | 18 / 20 Passed |