Content
96%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 exceptionally lean and rigorous skill body: executable commands with exact flags, precise input schemas, and a nine-step workflow with explicit freezing, quality-gate, and holdout checkpoints. The only structural weakness is minor bundle hygiene — one orphaned reference file, some dense inline input-format material, and unpathed cross-skill references.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is densely packed with non-obvious domain constraints ("Reviews are not calls, openings are not hires", "The miner... does not understand negation or generate synonyms") and teaches nothing Claude already knows; every line carries skill-specific rules, matching the lean anchor 5. | 5 / 5 |
Actionability | Provides copy-paste-ready commands with real flags ("python3 scripts/analyze_signals_v2.py --input accounts.csv --discover-phrases --max-phrases 500"), exact CSV schemas ("domain,status,website,jobs"), manifest fields ("analyzer_sha256", "config_sha256"), and JSON payload shapes, covering the common cases per anchor 5. | 5 / 5 |
Workflow Clarity | The 9-step workflow is clearly sequenced with explicit validation checkpoints for batch operations ("Pilot one or two rows and pass the quality gate... before scaling", "Freeze extraction rules, aliases, model and reference artifacts before validation", "preserve an untouched holdout"), satisfying the feedback-loop requirements of anchor 5. | 5 / 5 |
Progressive Disclosure | The task-to-reference routing table is excellent and all 19 referenced files exist one level deep, but references/pitfalls.md is an unreferenced orphan in the bundle, the CSV-format and source-adapter payload sections inline dense operational detail, and "deepline-gtm"/"deepline-plays" are named without paths, placing this just below anchor 5. | 4 / 5 |
Total | 19 / 20 Passed |