Content
53%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 well-conceived: a terse generic engine with a rule-key contract table, per-surface playbooks with concrete upstream-check commands, and appropriate delegation of the detailed workflow to a separate file. However, that referenced file (workflows/sweep.md) is absent from the bundle, leaving the core sweep procedure — the skill's central content — inaccessible, and verification steps named but not executable. The policy framework is strong; the delivered bundle is incomplete.
Suggestions
Ship workflows/sweep.md in the bundle (or inline the sweep phases into SKILL.md) so the referenced core procedure actually resolves — every path cited in the body should exist.
Make the Verify column executable: give the actual commands (e.g. the typecheck/test invocations, the workflow-run check) instead of naming suites generically.
Trim the duplication between the Operating policy table's auto-with-fixes row and the Playbooks rows (majors-included, same-PR fallout handling) so each rule is stated once.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence — no basic concepts are explained, and the tables pack contract details efficiently ("Rule keys are the contract; an inventory owner-rule cites them. Dropping or renaming a key changes the contract."). It is not 5 because the `auto-with-fixes` scope (majors included, fix fallout in the same PR) is stated in both the Operating policy table and again across several Playbooks rows, a redundancy that could be trimmed. | 4 / 5 |
Actionability | Some commands are concrete ("pnpm -r outdated", "cargo update --dry-run", "cargo audit", "kendex report") but the core procedure is delegated to [workflows/sweep.md](workflows/sweep.md), which does not exist in the bundle, and the Verify column names checks ("Typecheck + tests", "Workflow run") without executable commands. This matches 'Some concrete guidance but incomplete; missing key details'; it is not 4 because the gap is not minor — the primary workflow cannot be executed from what is written. | 3 / 5 |
Workflow Clarity | The sequence exists only as a named arrow chain ("inventory → detect → research → classify → upgrade-with-fixes → report the owner tier") with the actual step detail in the missing workflows/sweep.md, though checkpoints are partially present via the Verify column, test-suite gating for fork syncs, and "report only what actually failed, with error output". It is not 4 because a clear sequence with most checkpoints requires the steps themselves to be defined; as a batch operation touching many dependencies, the rubric also caps this dimension at 3 absent fully explicit validation. | 3 / 5 |
Progressive Disclosure | The body correctly delegates its workflow to a clearly signaled one-level-deep reference, but the bundle contains no workflows/, references/, scripts/, or assets/ directories — the sole referenced path (workflows/sweep.md) is dangling, so navigation to the primary content fails entirely, and 'orch SKILL.md § Harness-Safe Shell' is likewise unverifiable. This matches 'Minimal structure' as delivered; it is not 3 because the problem is not unclear signaling but a broken reference chain that makes the skill's main content inaccessible. | 2 / 5 |
Total | 12 / 20 Passed |