Content
75%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 operational skill: concrete ledger commands, explicit decision gates, a strict measurement/labeling protocol, and a useful rank-change quick-reference table. Main weaknesses are mild redundancy across sections and several references that point outside the skill's directory, plus the absence of an explicit refresh loop for NEEDS_REFRESH states.
Suggestions
Consolidate the duplicated source-label/evidence rules (stated in both Data Sources and Instructions) and the duplicate Tracking Setup Guide reference into a single authoritative section to tighten token use.
Add a short explicit procedure for handling NEEDS_REFRESH states (when to re-check, how to record the refreshed snapshot) to close the workflow's recovery loop.
Inline or verify the out-of-bundle paths (skill-contract.md, CONNECTORS.md, performance-monitor references, scripts/connectors/README.md) so every link in the body resolves from the skill's own directory.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~90-line body is dense and operational (tables, gates, commands) with no explanation of concepts Claude already knows; minor redundancy — evidence-label rules appear in both the Data Sources and Instructions sections, and the Tracking Setup Guide reference is listed twice — keeps it below lean. | 4 / 5 |
Actionability | Provides copy-paste-ready commands ("python3 \"${CLAUDE_PLUGIN_ROOT}/scripts/connectors/ledger.py\" record <domain> --source rank --data '{...}'" plus diff/trend variants), concrete decision-gate options, and a strict source-label taxonomy; gaps are that the ledger script and evidence profile are not in this bundle and the full output templates are deferred to a reference file. | 4 / 5 |
Workflow Clarity | Seven clearly sequenced steps with explicit checkpoints — a stop/ask gate with three enumerated options, a continue-silently gate, and a "Done when" completion checklist. Falls short of anchor 5 because NEEDS_REFRESH and conflict states are labeled but no explicit refresh/recovery loop procedure is given in the body. | 4 / 5 |
Progressive Disclosure | Good structure with well-signaled, one-level-deep local references (references/ranking-analysis-templates.md and references/tracking-setup-guide.md both exist), and bulk templates correctly split out of the body. However, four body references (../../../references/skill-contract.md, ../../../CONNECTORS.md, ../performance-monitor/SKILL.md, scripts/connectors/README.md) point outside the skill's own bundle and cannot be resolved within it. | 4 / 5 |
Total | 16 / 20 Passed |