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.
A tight, high-signal forcing-questions skill: concrete prompts, an exact positioning template, executable model commands, and a full output format, with zero wasted tokens. The two gaps keeping it from full marks are undefined verdict criteria in the workflow and external references that sit outside the skill's own bundle.
Suggestions
Define explicit criteria mapping the six answers to verdicts (e.g., RED if positioning fails the one-sentence test or any channel payback > 18 months; YELLOW if ICP cannot be named but positioning fits) so the verdict step is deterministic.
Add a validation checkpoint between answering and verdict (e.g., verify the positioning statement fits one sentence and each channel has CAC + payback numbers) with a retry loop for gaps.
Move the two model scripts into this skill's own scripts/ directory (or reference a resolvable path) so the Workflow commands execute from the skill's location.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: each question is one bolded prompt plus two terse bullets ("Persona ≠ ICP. ICP is real.", "If a channel's payback is > 18 months, it isn't a channel — it's a hobby."). There is no padding, no explanation of concepts Claude already knows, and every token instructs. | 5 / 5 |
Actionability | Guidance is fully executable: two copy-paste python commands in the Workflow, an exact one-sentence positioning template ("For [ICP], who needs [job], we are [category] that [differentiator] unlike [alternative]"), a concrete payback threshold (> 18 months), and a complete output format template. This is an instruction-only skill and its guidance is specific throughout. | 5 / 5 |
Workflow Clarity | The three-step sequence (run the models → answer the six questions in writing → apply the verdict) is clear, with the verdict as a decision checkpoint. It is not 5 because the boundaries between 🟢/🟡/🔴 are undefined (which question failures map to which verdict) and there is no validation or retry guidance if a model run fails or an answer cannot be produced. | 4 / 5 |
Progressive Disclosure | The single file is well organized (When to Run, Six Questions, Workflow, Output Format, Routing, Related) with nothing inlined that belongs in a separate file, and Routing/Related act as navigation. It is not 5 because the referenced materials point outside this skill's bundle via relative paths (e.g., "../../../c-level-advisor/skills/cmo-advisor/scripts/...", "../../agents/cs-cmo-advisor.md") that are not part of this skill's own references and cannot be resolved from it. | 4 / 5 |
Total | 18 / 20 Passed |