Content
61%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 skill is well-structured and token-efficient with good progressive disclosure, but its workflow and actionability suffer from bullet-form steps and a lack of executable command syntax for the rwb subcommands it names.
Suggestions
Provide concrete executable examples for the key rwb subcommands (e.g., the exact `rwb doctor` / `rwb run` invocation with expected arguments) to lift actionability.
Reformat the Workflow as a numbered sequence with inline validation checkpoints (confirm auth -> read-only probes -> validate -> escalate only if justified -> redact) so the validate-fix-retry loop is explicit.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean, bullet-driven, and assumes Claude's competence without explaining known concepts; a few placeholder bullets ("Outputs section", "Procedure section") carry low information but are not verbose, keeping it just shy of a 5. | 4 / 5 |
Actionability | It names concrete rwb subcommands (doctor, authorize, plan, run, summarize, manifest, validate, reconcile) but gives no executable command syntax or copy-paste examples, leaving guidance incomplete. | 3 / 5 |
Workflow Clarity | The Workflow section lists a sensible sequence as bullets and a separate Validation/Failure-Mode section adds fail-fast checkpoints, but the main workflow lacks inline numbered steps and explicit validate-fix-retry checkpoints, fitting the 3 anchor. | 3 / 5 |
Progressive Disclosure | A dedicated Progressive Disclosure section signals one-level-deep references to real bundle files (references/contract.yaml, evals.yaml, task-profile.json) plus deferred context, with only minor organization gaps from external Infrastructure paths. | 4 / 5 |
Total | 14 / 20 Passed |