Content
73%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 professionally structured strategy skill: a crisp contract (Reads/Writes/Promotes/Done when), a sequenced instruction set with completion gates, and disciplined routing to sibling skills. Its main weaknesses are systematic duplication of the scope guard and sibling handoffs (conciseness) and dependence on out-of-bundle files for key templates and referenced material (progressive disclosure, actionability).
Suggestions
State the scope guard once — the 'does not build the signup UX / write the confirmation emails / record the opt-in' exclusions appear in the intro, the Scope guard paragraph, and again after Instructions; keep a single canonical version.
Consolidate the sibling-skill handoffs: Reference Materials and Next Best Skill both enumerate consent-registry, landing-optimizer, and email-sequence-designer with overlapping descriptions — merge them into one section or cross-reference it.
Inline a minimal example of the 'operation: propose' consent.ndjson payload (and a short handoff-summary template) so the capture-flow spec step does not depend on an external skill-contract.md that may not resolve from the skill directory.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The scope exclusions ('It does not build the signup page, write the confirmation email, or record the opt-in') are stated three times — in the intro paragraph, the 'Scope guard' paragraph, and again after Instructions — and the sibling-skill handoffs (consent-registry, landing-optimizer, email-sequence-designer) are enumerated in both Reference Materials and Next Best Skill. The material is otherwise dense and free of concepts Claude already knows, but the systematic duplication is more than the 'minor instances' of the anchor-4 example, so it sits at 'mostly efficient but could be tightened'. | 3 / 5 |
Actionability | For an instruction-only planning skill the guidance is concrete: an exact save path ('memory/email/list-growth-designer/YYYY-MM-DD-<audience-or-goal>-growth-plan.md'), the specific consent-evidence items to capture ('timestamp, source, lawful basis, checkbox wording, IP/UA if used'), the propose mechanism via registry-events.py, and copy-paste Quick Start prompt templates. It is not 5 because the propose-request payload and the handoff-summary shape are deferred to an out-of-bundle skill-contract.md, leaving small execution gaps. | 4 / 5 |
Workflow Clarity | Eight clearly sequenced steps (confirm goal/jurisdiction → inventory → lead magnet → channels → capture flow → referral loop → metrics → compliance caveat) with an explicit 'Done when' checklist of four checkable criteria, missing-data handling ('Measured from the ESP export, or User-provided. Do not invent a baseline'), and human-in-the-loop checkpoints ('On user confirmation, save...', 'ask before writing'). It is not 4 because validation of completion and escalation/termination rules are explicit rather than implicit. | 5 / 5 |
Progressive Disclosure | The body is a well-organized overview with clearly signaled, purpose-labeled, one-level-deep references (Reference Materials lists each link with its role), and SEND/connector/security detail is appropriately pushed to separate files. It is not 5 because the bundle contains none of the referenced files — critical templates (save-results template, handoff summary format) live in an external ../../../references/skill-contract.md, and the deep relative paths (e.g., ../../../influencer/report/landing-optimizer/SKILL.md) are unverifiable from the skill directory. | 4 / 5 |
Total | 16 / 20 Passed |