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.
Excellent actionability and structure: executable gh commands for every path, a clear two-mode workflow with confirmation and pre-filter checkpoints, and clean single-file organization. The only trimmable fat is the rationale section's editorializing, and the batch backfill mode would benefit from a post-apply verification step.
Suggestions
Tighten the 'Why the timing is the whole point' section — the contamination argument can be stated in 2–3 sentences without the motivated-reasoning aside, saving ~10 lines.
Add a post-apply verification step to Mode B's batch pass (e.g., re-run the gh issue list filter after labeling to confirm no unrated issues remain), closing the workflow's one validation gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes competence — no generic-knowledge padding, every command shown once — but the 'Why the timing is the whole point' rationale and asides like 'That is motivated reasoning' could be tightened further. Not 5 because a few sentences of editorializing do not strictly earn their place; not 3 because there is no unnecessary concept explanation. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready commands cover the common cases: 'gh issue edit … --add-label "complexity:3"', the dual-label backfill command, the 'gh issue list … --jq' pre-filter for unrated issues, and the exclusion query. Mode A steps are concrete down to 'one sentence of justification naming the rubric row'. | 5 / 5 |
Workflow Clarity | Two clearly sequenced modes with real checkpoints — the owner confirm-or-amend loop in Mode A and the jq label pre-filter in Mode B — plus explicit handling of the unreachable-owner case. Not 5 because the batch backfill pass has no post-apply verification that labels actually landed; not 3 because validation is present (pre-filter, confirmation loop, unconfirmed marker), not absent. | 4 / 5 |
Progressive Disclosure | A self-contained single-file skill with well-organized sections (rationale, rubric, Mode A, Mode B, labels, consumers); nothing that belongs in a separate file is inlined, and the external pointers ('wiki/conventions/session-metrics.md', '@tools/pr-metrics') are one level deep and clearly contextualized as consumers rather than nested references. | 5 / 5 |
Total | 18 / 20 Passed |