Content
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill provides a well-structured advisory framework with a clear four-beat sequence, but it is significantly over-written for its purpose. Much of the content is motivational coaching language and meta-commentary that Claude doesn't need — the actual actionable content (the prompt template structure and the rules) could be delivered in roughly half the tokens. The cross-references to companion skills are a strength, but the lack of any validation/iteration mechanism and the monolithic presentation weaken it.
Suggestions
Cut the preamble and meta-commentary drastically — remove sentences like 'The best startup advisors don't just tell you what's wrong' and 'Most people describe their idea too small' which explain concepts Claude already understands. Lead directly with the prompt template.
Add a brief validation checkpoint between beats 1-2: e.g., 'Confirm the kernel resonates with the user before expanding to the bolder version' to create a feedback loop.
Extract the prompt template into a separate file (e.g., PROMPT.md) and keep SKILL.md as a concise overview with usage instructions and references, improving progressive disclosure.
Tighten the Rules section — several rules restate what's already implicit in the beat descriptions (e.g., 'Be specific when you push bigger' is already covered by the Beat 2 instructions).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Extremely verbose for what is essentially a prompt template. The preamble explains what good advisors do (Claude knows this), the prompt itself contains extensive coaching language and meta-commentary ('Most people describe their idea too small'), and the tips section adds more padding. The entire skill could be reduced to the prompt template with minimal framing. | 1 / 3 |
Actionability | The prompt template provides a structured four-beat framework with specific sub-questions and output format, which is reasonably actionable. However, it's a prompt template rather than executable code/commands, and some guidance remains abstract ('Be concrete enough to act on'). The landmine defusal format is well-structured but the overall approach is more descriptive than prescriptive. | 2 / 3 |
Workflow Clarity | The four beats are clearly sequenced and the ordering rationale is explicit ('if you lead with risks, people stop listening'). However, there are no validation checkpoints — no way to verify the kernel was correctly identified before pushing to the bolder version, no feedback loop if the user disagrees with the reframing. For an advisory/generative skill this is less critical, but the lack of any checkpoint or iteration step keeps it at 2. | 2 / 3 |
Progressive Disclosure | The content is a monolithic wall — the entire prompt template is inline with no separation of the template from supporting guidance. The tips section references other skills (yc-office-hours, pre-mortem, prioritize) which is good navigation, but the main body could benefit from separating the prompt template into its own file or at least having a concise overview before the full template. No bundle files exist to offload content to. | 2 / 3 |
Total | 7 / 12 Passed |