Content
88%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.
An unusually strong skill body: executable commands with exact flags, explicit error-recovery loops in the activation sequence, and disciplined conditional loading of one-level-deep reference files. The only deductions are a small duplicated recovery branch and a referenced customize.toml that is missing from the bundle.
Suggestions
Dedupe the 'script not found → offer bmad setup → re-run' recovery instructions in On Activation steps 1 and 3 by stating the recovery once and referencing it.
Ship a default customize.toml in the bundle (or reword the step-1 fallback so it does not point at a file that may not exist), since the fallback 'read {skill-root}/customize.toml directly with defaults' currently has no target.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence — no concept tutorials, no padding, every section instructs. Minor trimmable redundancy keeps it from 5: the 'Script not found: BMad is not set up here. Offer to run the bmad skill's setup...' recovery branch is repeated nearly verbatim in activation steps 1 and 3, and a few long parentheticals could be tightened. | 4 / 5 |
Actionability | Fully executable guidance throughout: exact commands with arguments ('uv run {project-root}/_bmad/scripts/resolve_customization.py --skill {skill-root} --project-root {project-root} --key workflow'), complete memlog invocations with every --field/--type/--by flag specified, exact file paths, and even the exact message to say to the user for the composer page. Not below 5 — nothing is pseudocode or left abstract. | 5 / 5 |
Workflow Clarity | Clear sequenced phases (On Activation 1-5, Run a Session, Choosing Techniques, Converging, Wrap-Up) with explicit validation and error-recovery loops: script-not-found branches that offer setup and re-run, 'if either hook list was non-empty, confirm every entry ran before continuing', and 'You can't see their browser, so never claim it opened'. This matches the anchor 'clear sequence with explicit validation steps; feedback loops for error recovery'. | 5 / 5 |
Progressive Disclosure | The body is a genuine overview with well-signaled, one-level-deep references loaded conditionally per mode (references/mode-*.md, converge.md, finalize.md, headless.md, resume.md), and all of those files plus scripts/brain.py and assets/brain-selector.html exist in the bundle. One minor gap keeps it from 5: {skill-root}/customize.toml is cited as the fallback to read when resolve_customization.py fails, but that file is absent from the bundle, so the stated fallback path is dangling. | 4 / 5 |
Total | 18 / 20 Passed |