Content
85%Reviews 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 multi-session skill with clear sequencing, validation checkpoints, and proper offloading of module templates to reference files. The main weakness is redundancy between the Anti-Patterns section and the protocols it restates.
Suggestions
Trim the Anti-Patterns entries that restate existing protocol rules (state-checking, one-question-at-a-time, saturation) and keep only net-new cautions such as the one-shot-session warning, to improve conciseness.
Tighten explanatory asides (e.g., the "so that no elicited knowledge is lost..." rationale in the opening) into imperative guidance.
Consider a brief "Module templates live in references/" pointer near the Module sequence table to make the file split explicit to the reader.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and actionable, but the Anti-Patterns section largely restates rules already in Session start protocol and Interview discipline ("Starting without reading state first", "Asking multiple questions at once", "Moving to Synthesis before saturation"), and some prose is explanatory padding. Not level 3 because those restatements and asides could be trimmed; not level 1 because the body is dense and largely earns its tokens rather than explaining concepts Claude already knows. | 2 / 3 |
Actionability | Provides literal, copy-ready guidance: the state.json schema as actual JSON, exact paths (modules/{moduleFile}, founders/{participant}.md), quoted projective prompts, and concrete validation rules (alphanumeric+hyphen participant names, 10–90 moduleFile enumeration, absolute outputPath). Not level 2 because guidance is fully executable rather than pseudocode or abstract direction. | 3 / 3 |
Workflow Clarity | Multi-step process is clearly sequenced (Session start protocol, Interview discipline, ordered Module sequence, State write protocol) with explicit checkpoints: the saturation signal, the per-module state checkpoint, and the terminal-module correctness guard that sets inProgressModule to null to avoid re-treating the finished brandbook as in-progress. Not level 2 because validation/checkpoint steps are explicit rather than implicit; not capped because file-write validation rules are present. | 3 / 3 |
Progressive Disclosure | SKILL.md is an overview; the eight detailed module templates live one level deep in references/ (10_purpose-why.md through 90_SYNTHESIS.md, all present and matching the module table), signaled via the Module sequence table and the State write protocol. Not level 2 because references are clearly signaled and content is appropriately split rather than inlined. | 3 / 3 |
Total | 11 / 12 Passed |