Content
67%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 policy-dense, well-structured instruction skill with concrete regexes, commands, and mapping tables, and a sequenced workflow with real validation checkpoints. Its weaknesses are verbosity — re-explaining conventional commit format and cramming multi-condition rules into run-on prose — and an absence of explicit error-recovery loops in the workflow.
Suggestions
Cut the conventional commit format breakdown ("type — required", "(scope) — optional", "! — optional") and trim the 11-example list to 2-3 representative ones; keep only the project-specific type-to-label mapping, since Claude already knows conventional commit syntax.
Break the dense run-on rule sentences in Critical Rules and Commands into short bullets, one condition per bullet, to preserve the same constraints in fewer, more scannable tokens.
Add explicit error-recovery branches to the workflow (e.g., "if the linked issue lacks status:approved or a required check fails, stop and report — do not proceed to PR creation") to close the feedback-loop gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient — the tables and rule lists carry genuinely project-specific policy — but there is unnecessary explanation of conventional commit format (the "type — required...", "(scope) — optional", "! — optional" breakdown and 11 examples re-teach a format Claude already knows), and dense run-on policy sentences like "Baseline attribution requires reproducing the same failing command/environment on a comparable isolated clean base..." could be tightened into bullets. It is not a 4 because multiple sections (commit format explanation, full example list, 11-row branch table) could be trimmed; it is not a 2 because most content is novel policy, not padding or known-concept explanation. | 3 / 5 |
Actionability | Concrete, executable guidance dominates: exact regexes for branch names and commit messages, precise commands ("go test ./...", "go run ./internal/gofmtcheck", "cd e2e && ./docker-test.sh"), and complete type-to-label mapping tables. It falls short of fully copy-paste-ready because the actual branch/PR creation commands are deliberately absent ("Do not assume main or execute branch/remote mutations from examples") and the PR body section is explicitly schematic placeholders, leaving minor gaps the model must fill at runtime. | 4 / 5 |
Workflow Clarity | The 6-step workflow is clearly sequenced with most checkpoints present: verify "status:approved" before proceeding, "run applicable local checks and report failures honestly", "mark checkboxes only after readback", and "Read target policy and status to identify REQUIRED checks; do not infer requiredness from a pending optional run", plus a contributor checklist and automated-checks table. It is not a 5 because error-recovery feedback loops (what to do when a check fails or the issue lacks approval) are implicit rather than explicit validate→fix→retry steps. | 4 / 5 |
Progressive Disclosure | The ~190-line single-file body has good section structure (When to Use, Critical Rules, Workflow, Branch Naming, PR Body Format, Automated Checks, Conventional Commits, Commands) with clearly signaled one-level-deep pointers to repo files (CONTRIBUTING.md, internal/assets/skills/issue-creation/SKILL.md, .github/PULL_REQUEST_TEMPLATE.md). It is not a 5 because there is no bundle at all: reference material like the commit example list and type-mapping tables sits inline in one long file, and the under-50-lines exception does not apply. | 4 / 5 |
Total | 15 / 20 Passed |