Content
70%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.
The body delivers an exemplary workflow: ordered checklist, explicit user-approval validation gates, self-review loop, and concrete verbatim scripts for key moments. Its weaknesses are repetition of the hard-gate rule, an overlong Visual Companion section, and a progressive-disclosure gap where the referenced visual-companion.md is missing from the bundle and the bundled scripts are never mentioned.
Suggestions
State the hard gate once (in the HARD-GATE block) and reduce the other three restatements to brief pointers, trimming the Visual Companion section to the offer script plus the browser-vs-terminal test.
Fix the dangling reference to skills/brainstorming/visual-companion.md — either include the file in the bundle or reference the actual bundled scripts (scripts/server.cjs, scripts/start-server.sh) so the detail layer is real and discoverable.
Add one compact example of a proposed approach with trade-offs and a recommendation to make the 'Propose 2-3 approaches' step concretely executable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The no-implementation gate is repeated four times ("Do NOT invoke any implementation skill... until you have presented a design", the anti-pattern section, "The ONLY skill you invoke after brainstorming is writing-plans", and "Do NOT invoke any other skill"), and the Visual Companion section restates the browser-vs-terminal test with worked examples. Mostly efficient, but these padded/redundant passages could be trimmed — anchor 3 rather than 4's 'minor instances'. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is largely executable: an ordered 9-item checklist, an exact spec path (`docs/superpowers/specs/YYYY-MM-DD-<topic>-design.md`), verbatim messages for the companion offer and the user review gate, and a concrete 4-step self-review procedure. Minor gaps remain — the "invoke writing-plans skill" step depends on an external skill, and no example shows what a proposed approach or design section looks like. | 4 / 5 |
Workflow Clarity | The 9-item checklist gives a clear ordered sequence, and validation checkpoints with feedback loops are explicit throughout: "get user approval after each section", the self-review fix-inline loop, and the user review gate with a 'changes requested' -> revise cycle — also visualized in the dot graph. This matches anchor 5 (explicit validation steps, feedback loops, checklists). | 5 / 5 |
Progressive Disclosure | The body appropriately defers detail with a clearly signaled one-level reference ("read the detailed guide before proceeding: skills/brainstorming/visual-companion.md"), but that file is not present in the bundle (no references/ or skills/ directory exists), and none of the five scripts/ bundle files (server.cjs, helper.js, start-server.sh, etc.) are referenced or discoverable from the body. Structure is otherwise good, but the dangling reference and unlisted bundle content leave navigation incomplete — anchor 3. | 3 / 5 |
Total | 15 / 20 Passed |