Content
63%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 dense, well-structured operational skill with concrete thresholds, clear phase-gated workflow, and mandatory checkpoints \u2014 genuinely actionable for a research-design agent. Its weaknesses are duplication (the capabilities comment block restates two later sections) and a progressive-disclosure architecture whose referenced files are absent from the bundle, leaving the routing tables pointing at nothing.
Suggestions
Remove the CAPABILITIES_SUMMARY / COLLABORATION_PATTERNS / BIDIRECTIONAL_PARTNERS HTML comment block \u2014 it duplicates the Trigger Guidance and Collaboration sections nearly verbatim and costs ~25 lines of context.
Ship the referenced 'reference/*.md' and '_common/*.md' files in the bundle, or inline the load-bearing content (e.g. the bias checklist, secondary thresholds) so the Workflow, Recipes, and Reference Map tables don't point to missing files.
Consolidate the duplicated routing guidance \u2014 the 'Route elsewhere' list in Trigger Guidance and the 'Route out' note at the end of Output Routing say the same thing and could be merged.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body assumes Claude's competence (no concept explanations, dense rule-based content), but it is noticeably duplicated: the CAPABILITIES_SUMMARY/COLLABORATION_PATTERNS/BIDIRECTIONAL_PARTNERS comment block (lines 6-31) restates the Trigger Guidance and Collaboration sections almost entirely, route-out rules appear in both Trigger Guidance and Output Routing, and the Reference Map repeats the 'Read' columns of the Workflow and Recipes tables. This fits anchor 3's 'mostly efficient but could be tightened' rather than 2, since most sections carry operational rules rather than padding. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is highly concrete and executable: exact thresholds ('SUS >68', 'Cronbach's \u03b1 \u2265 0.70', 'qual usability <5 users, quant <30'), a canonical output structure ('## User Research Report \u2192 ### Research Objective \u2192 ...'), specific data-governance requirements ('SOC 2 Type II, GDPR readiness with a DPA'), and per-phase required actions. Minor gaps remain because some phase actions are high-level ('Code data, identify patterns, check bias') with the executable detail delegated to reference files. | 4 / 5 |
Workflow Clarity | The DEFINE \u2192 DESIGN \u2192 ANALYZE \u2192 SYNTHESIZE \u2192 HANDOFF workflow is a clear table with required action, key rule, and read-target per phase, and mandatory checkpoints exist ('Run bias checks in design, execution, analysis', '\'analysis\' requires a bias check', 'Include confidence and limitations'). It matches anchor 4's 'clear sequence with most checkpoints present' but not anchor 5, which expects explicit error-recovery feedback loops; the destructive/batch cap does not apply since this is a research-design skill. | 4 / 5 |
Progressive Disclosure | The design intent is strong \u2014 one-level-deep, clearly signaled references with a Reference Map ('Read this when') and per-recipe 'Read First' lists \u2014 but the bundle contains only SKILL.md: none of the ~21 referenced 'reference/*.md' or '_common/*.md' files exist, so per the guideline to score against the actual bundle structure, navigation leads to missing files and the split content is not delivered. This places it at anchor 3: good structure within the file, but the progressive-disclosure promise is unfulfilled. | 3 / 5 |
Total | 14 / 20 Passed |