Content
85%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 well-structured instruction skill: the 8-step workflow is concrete, validated, and anchored by real one-level-deep reference modules with clear use conditions. The only weakness is redundancy — three sections restate the procedure and rules, which inflates token cost without adding guidance.
Suggestions
Merge "Core Function" and "What This Skill Should Not Do" into the 8-step execution section and Hard Rules, removing their duplicated restatements of the same points.
Cut the "Quality Standard" closing section — its strong/weak output definitions repeat the steps and Hard Rules verbatim in spirit.
State the never-fabricate rule once (e.g., in Hard Rules plus references/literature-integrity-rules.md) instead of repeating it in the description, section H, and Hard Rule 10.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — tight bulleted lists, no explanations of concepts Claude already knows — but it could be tightened: "Core Function", "What This Skill Should Not Do", and "Quality Standard" largely restate the 8-step procedure and Hard Rules, and the separate-observation-from-interpretation and never-fabricate rules each appear 3–4 times across sections. This matches anchor 3 (mostly efficient with some unnecessary repetition) better than anchor 2, since no section is pure padding and the core procedure text earns its place. | 3 / 5 |
Actionability | As an instruction-only skill, the guidance is fully concrete: an ordered 8-step procedure with named classification schemes (four support-strength categories, five overinterpretation patterns, five figure types), a defined figure-to-claim table format in references/figure-to-claim-framework.md, and a mandatory output structure (sections A–H) with specified fields. Specific examples cover the common cases (quick scan vs. full audit, captions-only material, missing figures), matching the top anchor for concrete, executable guidance. | 5 / 5 |
Workflow Clarity | The 8 steps are explicitly ordered ("always run in order") with explicit validation checkpoints: Step 1 requires declaring missing figures before judging support, Step 8 is a dedicated self-critical final check (strongest link, weakest link, biggest overinterpretation risk), and the output's Risk Review section provides a final audit. The 15-item Hard Rules checklist covers edge cases like ambiguous figure-to-claim links and unreadable panels. This matches the anchor for clear sequence with explicit validation steps and checklists. | 5 / 5 |
Progressive Disclosure | The "Reference Module Integration" section clearly signals all 7 one-level-deep references with per-file use conditions ("Use for mapping each figure...", "Use when the visual evidence is weaker..."); all 7 referenced files exist in references/ and none nest further references. The body appropriately keeps the overview (steps, output structure, rules) while details live in the reference files, matching the anchor for a clear overview with well-signaled one-level-deep references. | 5 / 5 |
Total | 18 / 20 Passed |