Content
81%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 a highly actionable, well-sequenced coordinator workflow with strong validation gates on its one destructive operation (merging), and it splits variant-specific and deterministic content appropriately into one-level-deep references. Its main weakness is redundancy: the Common Mistakes, Red Flags, and Important sections restate the same rules two to three times, inflating token cost.
Suggestions
Merge the "Common Mistakes", "Red Flags", and "Important" sections into a single table — each currently restates the same rules (links-only graphs, unverified "done" claims, hand-topo-sorting, per-dimension trust, staying resident) two to three times, costing tokens without adding information.
Verify and ship the referenced sibling files (`github-mode.md`, `merge-gate.md`, `requirements.json`, `lib/schedule.py`, `lib/gate.py`) alongside SKILL.md — none are present in this bundle, so the body's point-of-need references cannot currently be resolved.
Trim the historical aside in Phase B ("Real case: the MES-4414 tag tickets...") to one clause; the operational rule (always parse both sources) stands on its own.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The instruction is domain-specific and mostly efficient, but "Common Mistakes", "Red Flags — STOP if you think", and "Important" each repeat the same rules nearly verbatim (empty issuelinks, verify "done" claims, don't eyeball the topo sort, per-dimension trust, stay resident — each stated 2-3 times). This matches the anchor for mostly efficient content that could be tightened rather than only minor trimming. | 3 / 5 |
Actionability | Copy-paste-ready commands with plugin-root fallback (`python3 "$gate" mode "<epic ref>"`), an exact scheduler input JSON shape, exact `spawn_task` field values, a per-ticket brief template, and exit-code handling (0/10/1) — fully executable guidance covering the common cases, matching the top anchor. | 5 / 5 |
Workflow Clarity | Phases A-F are clearly sequenced with a mermaid flowchart and explicit validation checkpoints at every risky point: re-verify Jira + MR before advancing a wave, abstain-and-ask on dependency cycles, `gate.py premerge` and a `REVIEW:` verdict at the head SHA before the destructive merge step — matching the anchor for clear sequence with explicit validation and error-recovery feedback loops. | 5 / 5 |
Progressive Disclosure | Structure is good: variant detail is deferred to one-level-deep siblings ("read the sibling `github-mode.md`", `merge-gate.md` before Phase E) and deterministic logic to `lib/schedule.py` / `lib/gate.py`, each clearly signaled at point of need and gated on mode output. Not top marks because the referenced sibling files are not present in this bundle to verify, and the three redundant mistake/red-flag/important sections inline content that should be consolidated or externalized. | 4 / 5 |
Total | 17 / 20 Passed |