Content
86%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 is an exemplary dispatcher: lean, dense with non-obvious operational rules, and routing every situation to a real, one-level-deep reference file. Its only ceiling is that executable detail mostly lives in the bundle files rather than the body itself, which is the appropriate design for this skill.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every one of the ~30 lines is an operational fact Claude cannot infer — the YOLO-mode check, the file layout tree, and eight terse gotchas — with nothing explaining known concepts; matches 'lean and efficient; every token earns its place'. | 5 / 5 |
Actionability | Includes a copy-paste command ('[ "${AGENTTEAMS_YOLO:-}" = "1" ] || [ -f ~/yolo-mode ]'), exact paths ('shared/tasks/{task-id}/', 'meta.json'), and unambiguous rules ('Do NOT proceed to next phase while REVISION_NEEDED is pending'), but most executable procedures are delegated to the reference docs rather than shown in the body — anchor 4, not 5's fully self-contained coverage. | 4 / 5 |
Workflow Clarity | Routing is unambiguous via the situation-to-doc table and includes explicit checkpoints ('Check YOLO mode BEFORE drafting the plan', the REVISION_NEEDED gate, the mandatory 'All tasks complete' step), but the end-to-end sequences themselves live one level deep in the references, leaving minor in-body sequence gaps — anchor 4 rather than 5. | 4 / 5 |
Progressive Disclosure | A clear overview with well-signaled, verified one-level-deep references: the table maps every situation to one of four existing reference files, explicitly instructs 'Read the relevant doc before executing. Do not load all of them', and the scripts are invoked with full paths from within the references — matches the anchor-5 example pattern. | 5 / 5 |
Total | 18 / 20 Passed |