Conciseness

The body is dense and assumes Claude's competence — no basic-concept explanations, numeric thresholds instead of prose — but the 6-line "Fontes" provenance section (including the time-sensitive "set/2026") and partial overlap between "Anti-Padrões" and the flow/tactics rules are minor instances that could be trimmed. Not score 5 because of those few tokens that do not earn their place operationally; not score 3 because the padding is minor, not 'some unnecessary explanation'.

Actionability

Concrete instruction-only guidance: numeric thresholds ("TTV p50 dos ativados abaixo de 5 minutos", "No máximo 5 perguntas"), named events (`personalized_home_viewed`, `habit_loop_completed`, `streak_at_risk`), and a defined delivery-table format with columns. Not score 5 because there is no worked example (e.g., a filled delivery-table row or sample copy) covering a common case; not score 3 because the guidance is executable, not pseudocode-level.

Workflow Clarity

The 8-step "Fluxo Obrigatório" is clearly sequenced and ends in an explicit validation checkpoint ("Checkpoint antes de entregar: todo movimento tem evento nomeado; nenhuma métrica fora das faixas citadas sem `HIPOTESE`; nenhum exemplo externo sem `ANALOGIA`"), reinforced by a separate "Evidência de Conclusão" checklist. Not score 4 because checkpoints are explicit and checklists are present, matching the anchor-5 pattern.

Progressive Disclosure

The body is a true overview that keeps each tactic as a one-line ordering item and delegates patterns, rules, movements, and events to the single verified, one-level-deep `references/tactics.md`, clearly signaled both in the flow ("Auditar as 5 táticas com `references/tactics.md`") and in the "Recursos" section. Not score 4 because the split is clean and navigation is unambiguous with no inlined content that belongs in the reference.

Specificity

"Diagnostica retencao e projeta loops de uso recorrente" names the domain and exactly two concrete actions, matching the '1-2 concrete actions, but not comprehensive' anchor; it omits other capabilities the skill actually delivers (tactic audit, retention-event definition, D1/D7/D30 scorecard, 4-week plan).

Completeness

Both questions are answered explicitly: the 'what' ('Diagnostica retencao e projeta loops de uso recorrente') and a fully explicit 'when' with concrete trigger phrases ('Trigger em: "retenção", "retenção baixa", ...'). Not score 4 because the 'when' clause is as explicit and specific as the anchor-5 example.

Trigger Term Quality

Eighteen natural triggers including colloquial user phrasings ("deixar o app grudento", "usuário some depois do primeiro dia"), metric names ("D1", "D7", "D30"), and synonyms ("hábito", "loop de hábito", "streak", "engajamento recorrente") give comprehensive coverage; not score 4 because no common variation is obviously missing.

Distinctiveness Conflict Risk

The retention niche is clear with distinctive triggers ("switching cost", "lock-in ético", "streak", "churn de uso"), but broader terms like "personalização do app", "comunidade no app" and "engajamento recorrente" create minor overlap risk with adjacent UX/growth skills. Not score 5 because those generic terms could plausibly fire for neighboring skills.