Conciseness

Quotes: 'Reviews are not calls, openings are not hires', 'It does not understand negation or generate synonyms', 'The script cannot verify timestamp provenance' — the body is telegraphic and dense, assumes full competence, and contains no padding or explanations of concepts Claude already knows. Every line carries operational information, matching the lean/efficient anchor exactly rather than the 4 anchor's 'minor instances of over-explanation'.

Actionability

Quotes: 'deepline auth register --wait auto', 'python3 scripts/analyze_signals_v2.py --input accounts.csv --discover-phrases --max-phrases 500 --min-phrase-accounts 2 --output candidates.json', 'CSV columns: domain,status,website,jobs', adapter JSON shapes — commands are copy-paste ready with full flags, and input schemas, config formats, and output fields are all concretely specified.

Workflow Clarity

Quotes: 'Pilot one or two rows and pass the quality gate on the generated outputs before scaling', 'Check collection with the quality gate... before citing lift', 'Freeze extraction rules, aliases, model and reference artifacts before validation', 'Report truncation and increase the cap when needed' — an 8-step sequenced workflow with explicit validation checkpoints, gating before paid scaling, freeze-before-validation discipline, and a retry loop, matching the anchor with explicit validation and error-recovery feedback.

Progressive Disclosure

Quotes: the task→reference table ('Find phrases and buyer language | Keyword catalog...') plus inline links to quality-gate.md, scoring-pitfalls.md, signal-interpretation.md, and report-template.md — all referenced files exist in references/ and are one level deep. Not 5 because operational detail like source-adapter JSON shapes, manifest contents, and statistical methodology (Wilson intervals, Fisher p-values) is inlined in SKILL.md where it could be split into a reference, a minor organization gap versus the 'content appropriately split' anchor.

Specificity

Quotes: 'discovering niche signals, auditing ICP or won/lost evidence, rescoring accounts, or building account and lead scoring Plays' — several concrete domain-specific actions are named. Not 5 because the actions live inside the when-clause and a couple ('discovering niche signals') are mildly abstract, leaving minor coverage gaps versus the comprehensive multi-action anchor.

Completeness

Quotes: 'Use when discovering niche signals... or building account and lead scoring Plays' — both a what (building/auditing/rescoring scoring artifacts) and an explicit when are present, plus a skip clause ('Skip pure outreach copy or contributor skill installation tasks'). Not 5 because the 'what' is only expressed as gerunds inside the when-clause rather than a standalone declarative capability statement, so it could be more explicit.

Trigger Term Quality

Quotes: 'Triggers on fit scoring, engagement scoring, external proxies, and scoring leakage' — good natural keyword coverage a scoring buyer would say, plus 'rescoring accounts' and 'account and lead scoring Plays'. Not 5 because common synonyms like 'lead scoring model', 'ICP scoring', or file-format/extension equivalents are missing; clearly above the 3 anchor's 'missing common variations'.

Distinctiveness Conflict Risk

Quotes: 'auditing ICP or won/lost evidence', 'building account and lead scoring Plays', 'Skip pure outreach copy or contributor skill installation tasks' — a clear niche with distinct triggers and an explicit skip clause that actively reduces overlap with adjacent outreach/sales skills, matching the minimal-conflict anchor.