Content
75%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, highly actionable triage workflow: clear sequencing, concrete thresholds, checklists, and an explicit report format. The main weakness is redundancy — screening criteria are restated across the checklist, classification bands, and issues playbook — which inflates token cost without adding guidance, alongside a missing final validation step and inlined reference material.
Suggestions
Deduplicate the screening standards: the classification bands in Step 4 and the 'Common NDA Issues' section restate criteria already covered by the Step 3 checklist (carveouts, residuals, non-solicitation, term lengths) — reference the checklist items by name instead of restating them to cut roughly a third of the body.
Move 'Common NDA Issues and Standard Positions' into a one-level reference file (e.g., references/issues.md) loaded only when issues are actually found, keeping SKILL.md as the overview plus core workflow.
Add an explicit validation step before classification: re-verify each GREEN criterion against the completed screening table (and confirm every report field traces to the NDA text or 'Not stated') to catch mismatches between the screening results and the assigned classification.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is substantive domain content rather than explanations of things Claude already knows, but the same standards are stated two or three times: the carveout list appears in 'Standard Carveouts', again in the GREEN 'All of the following must be true' criteria, and again under 'Common NDA Issues'; the residuals standard appears three times ('No residuals clause (or narrowly scoped)', 'Residuals clause present but narrowly scoped', 'Issue: Broad Residuals Clause'). This is more than 'minor instances' (anchor 4) but not padded with unnecessary explanation (anchor 2), so it fits 'mostly efficient but could be tightened'. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is fully executable: concrete numeric thresholds ('1-3 years, survival 2-5 years', '10+ years, or perpetual without trade secret justification'), a copy-paste-ready triage report template, a routing table with timelines, and specific redline approaches ('limit to targeted solicitation... and set a short term (12 months)'). Specific examples cover the common cases, matching the top anchor. | 5 / 5 |
Workflow Clarity | A clear six-step sequence (accept → load playbook → screen → classify → report → route) with per-criterion checklists and explicit ambiguity handling ('choose the more cautious classification and say which fact was missing', 'Write Not stated... do not infer'). It is not 5 because there is no final output-validation step — e.g., re-verifying that every GREEN criterion is satisfied by the screening table before assigning GREEN — which is the missing checkpoint anchor 4 ('minor validation gaps') describes; not 3, since checkpoints for ambiguous and missing facts are explicit and the operation is read-only analysis, so the destructive/batch cap does not apply. | 4 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are absent) and the single SKILL.md is well-organized with clear section headers; the checklist, classification bands, and report template are needed on every run, so inline placement is appropriate. It is not 5 because the ~30-line 'Common NDA Issues and Standard Positions' section is lookup material that could live in a one-level reference file, and at ~250 lines the skill exceeds the simple-skill exception; not 3, because there are no buried or nested references and navigation is easy. | 4 / 5 |
Total | 16 / 20 Passed |