Content
76%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 tight, executable instruction set for an instruction-only skill, with concrete commands and a workable rubric. The main gap is the batch backfill path (Mode B), which applies labels across up to 200 issues with no validation or verification feedback loop.
Suggestions
Add a validation step to Mode B's batch pass: after the one-pass labeling, re-run the jq filter (or `gh issue list --label "complexity:unconfirmed"`) to confirm no unrated issues remain, so misapplied or skipped labels are caught.
Add a checkpoint in Mode A after applying the label — e.g. verify the label took (the repo actually has the `complexity:*` labels) before carrying on — so a failed `gh issue edit` doesn't silently drop the rating.
Trim the rhetorical passages in the timing section ("the number gets talked into agreeing with the spend…") to a single declarative rule statement to tighten conciseness.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Efficient and project-specific throughout (the timing rationale, the Fibonacci rubric, the label semantics), with only minor rhetorical passages that could be trimmed ("the number gets talked into agreeing with the spend, and the metric quietly starts confirming whatever already happened") and a trailing "What consumes this" context section that is partly dispensable. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready commands cover the common cases: `gh issue edit 412 --repo englishstreetventures/osn --add-label "complexity:3"` for both modes, the jq-filtered `gh issue list` command for the batch pass, and the exclusion query — plus a concrete five-row rubric table. | 5 / 5 |
Workflow Clarity | Mode A has a clear four-step sequence with an unreachable-owner fallback, but Mode B is a batch operation (a one-pass labeling over up to 200 issues) with no verification step — no post-check that labels applied or re-run of the filter to confirm completion. Per the batch-operations cap, workflow clarity cannot exceed 3. | 3 / 5 |
Progressive Disclosure | No bundle files exist, and the 104-line body is well organized into purposeful sections with all content appropriately inline and one-level pointers to external consumers ("wiki/conventions/session-metrics.md", "@tools/pr-metrics"). It exceeds the ~50-line simple-skill threshold, so it stops short of the 5 anchor's clean overview-plus-references shape. | 4 / 5 |
Total | 16 / 20 Passed |