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.
A high-quality orchestration body: fully executable commands with exact payloads, a clearly sequenced loop with validation checkpoints, error-recovery feedback loops, and precise hand-back contracts. The only deductions are minor trimmable rationale passages with dated detail and inline content that could move to reference files.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense operational prose where nearly every sentence carries commands, states, or refusal handling, but a few spots could be trimmed: the headless-session rationale embeds a dated version record ("recorded on 2026-10-24 with Claude Code 2.1.310") and Step 2 explains dirty-repo mechanics partly inferable from context. | 4 / 5 |
Actionability | Fully executable throughout: exact CLI invocations with complete JSON payloads (status/worktree/merge/implement new/spec amend), exact first-line hand-back markers (DONE:/QUESTION:/FAILED:), and concrete handling for named error codes like epic_merge_conflict and merge_lock_timeout. | 5 / 5 |
Workflow Clarity | Steps 1–8 form a clear, explicitly looped sequence with validation checkpoints and feedback loops: refuse-before-start on epic.run.problems, merge retry on merge_lock_timeout, cut-off child detection and escalation, stopping mode, and a final-report checklist — thorough validation for a batch operation. | 5 / 5 |
Progressive Disclosure | With no bundle files present, the body's own structure scores well: numbered steps, bold headers, and an explicit loop map with nothing buried or nested. Minor gaps: sections like the child-prompt template and headless/trace-span guidance are long enough to live in a reference file, and at ~135 lines the skill exceeds the sub-50-line simple-skill exception. | 4 / 5 |
Total | 18 / 20 Passed |