Content
88%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.
An exemplary orchestration-hub skill: every section is dense, actionable, and validated, with details correctly pushed into 9 conditionally-loaded reference files and 2 executable scripts. The only notable costs are some cross-section repetition of authorization/re-read rules and sibling references cited from other reference files.
Suggestions
Consolidate the authorization-scope rules stated in both the Outcome Contract and Autofix Routing into one canonical section referenced from the other, trimming the duplicated sentences.
Have mode files reference the shared procedure sections by the SKILL.md anchor (e.g. "the Hard Stops in the parent") rather than citing sibling reference files directly, keeping reference chains strictly one level deep.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense imperative rules throughout with zero explanation of concepts Claude already knows, assuming competence. However, authorization scope is stated in both the Outcome Contract and Autofix Routing, re-read rules appear in both Hard Rules and Get the Diff, and some run-on rule sentences could be trimmed — anchor 4 (efficient, minor instances that could be trimmed). | 4 / 5 |
Actionability | Copy-paste-ready commands throughout (git status --short --branch -uall, gh run view <id> --json status,conclusion, git fetch origin pull/<n>/head:refs/tmp/pr-<n>, git merge-tree, bash <skill-base-dir>/scripts/run-tests.sh, the git worktree add --detach isolation recipe) plus concrete thresholds, routing-class tables, and a complete sign-off template with exact field values. Fully executable guidance covering the common cases. | 5 / 5 |
Workflow Clarity | A clear sequence (worktree safety preflight, mode picker, context extraction, baseline freeze, scope, checks, verification, sign-off) with explicit validation checkpoints and feedback loops: verdict expiry and re-read rules after fixes, re-run the exposing check, regression-test-before-fix, isolated verification in a clean worktree, and CI state confirmation before merge. Anchor 5: explicit validation steps, feedback loops, and checklists for a complex process. | 5 / 5 |
Progressive Disclosure | Conditional mode loading via an intent table, and all ten cited bundle paths (9 references + scripts/run-tests.sh) resolve to real files with flat, non-nested content. Minor gap: mode files cross-cite sibling references (mode-ship.md points to project-context.md, mode-triage.md to public-reply.md), creating a mild second hop that keeps it below anchor 5's clean one-level-deep picture. | 4 / 5 |
Total | 18 / 20 Passed |