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.
The body is a tight, well-structured read-only workflow with executable commands, an explicit verification checkpoint, fallback handling, and a closing checkpoint section. Actionability and progressive disclosure sit just below the top because $VAULT/$QUERY and PRODUCT_ROOT are resolved narratively and the referenced files are sibling-skill/external paths with no local bundle to verify against.
Suggestions
Add a short concrete snippet that resolves $VAULT and $QUERY from the user's selection so the retrieve command is copy-paste-ready end to end, not just the invocation line.
Clarify that provenance.md and the sibling SKILL.md references live under the installed product root (consistent with the PRODUCT_ROOT note) so the navigation target is unambiguous.
Consider a one-line inline summary of the claim-ledger schema fields (accepted/provisional/contested/unsupported) or confirm provenance.md covers them, so the Assess-evidence step is self-contained without the external reference.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and efficient with no padding or explanation of concepts Claude already knows; every section (security, root resolution, depth, retrieve, assess, answer, checkpoint) earns its place and assumes Claude's competence. | 5 / 5 |
Actionability | Provides concrete executable bash commands (the contracts --verify check and retrieve invocation with explicit flags) plus specific ledger paths, but $VAULT/$QUERY resolution and the PRODUCT_ROOT placeholder are described narratively rather than as a fully copy-paste-ready sequence. | 4 / 5 |
Workflow Clarity | Five-step Retrieve sequence with an explicit validation checkpoint (capability --verify), a named fallback path ("fall back to wiki/index.md... Say which fallback was used"), and a Checkpoint section that re-verifies citations and names evidence gaps; the skill is explicitly read-only so the destructive/batch cap does not apply. | 5 / 5 |
Progressive Disclosure | Well-organized sections with clearly signaled one-level-deep references (provenance.md, sibling wiki-retrieve and wiki-cli SKILL.md files), but the referenced paths point to sibling skills / a non-local references directory rather than bundle files within this skill, so navigation cannot be fully verified against an actual bundle structure. | 4 / 5 |
Total | 18 / 20 Passed |