Content
85%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 well-engineered, highly actionable workflow skill with clear sequencing, an explicit approval gate, and clean one-level-deep reference splitting. Its one weakness is conciseness: the approval-gate discipline and anchoring rationale are restated several times and could be tightened.
Suggestions
Consolidate the approval-gate guidance: step 3, the pre-gate protocol, the amendments section, and Boundaries all re-explain the same gate — state the rule once authoritatively and reference it, rather than re-deriving it each time.
Trim the meta-commentary passages (e.g. "The documented failure here: composing the plan in thinking...") to the actionable instruction; the failure mode can be named without the narrative.
Tighten the anchoring rationale in step 2 — the core point (triage in a fresh subagent, withhold session reasoning) can be made in roughly half the length without losing the why.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and free of basic-concept padding, but the approval gate is re-explained across step 3, the pre-gate protocol, boundaries, and amendments, and the anchoring rationale and "documented failure" meta-commentary could be tightened without losing the guidance. | 3 / 5 |
Actionability | Gives concrete commands ("gh auth status", "git remote get-url origin"), exact tool calls (subagent_type: "general-purpose", not "fork"), exact file paths (review/pr-<N>-comments-plan.md), a category decision table, and a commit-subject example — fully actionable for an instruction-only skill. | 5 / 5 |
Workflow Clarity | A 6-step numbered sequence with explicit validation (the results gate), feedback loops (back with sources / proceed / change, re-present the revised plan), and deliberate handling of batch review items — checkpoints and error recovery are explicit. | 5 / 5 |
Progressive Disclosure | SKILL.md is a platform-agnostic overview; platform-specific commands are split into real, one-level-deep references (github.md, gitlab.md, local.md) that are all linked from the body and cross-linked to each other, making navigation easy. | 5 / 5 |
Total | 18 / 20 Passed |