Lean, dense prose with no padding — it never explains concepts Claude already knows and every line carries instruction. Not 2 because there is no unnecessary explanation to tighten away.

Provides executable commands ('uv run ... list_customizable_skills.py', the resolver invocation), exact file paths, and precise TOML merge semantics by field type. Not 2 because the guidance is concrete and copy-ready rather than pseudocode.

A clear numbered sequence (Preflight → Steps 1–6) with explicit validation in Step 6 and a feedback loop ('re-enter Step 4 with the verify output as context'). Not 2 because validation checkpoints and error-recovery loops are explicit, not implicit.

Well-organized sections with one-level references to a verified bundled script (scripts/list_customizable_skills.py exists) and a single external guide URL, no nested reference chains. Not 2 because references are clearly signaled and appropriately split rather than inlined.

Names a concrete artifact (customization overrides for installed BMad skills) and two specific actions ('Authors and updates'), avoiding vague filler. Not below 3 since it states concrete actions rather than abstract language like 'helps with'.

Explicitly answers both what it does ('Authors and updates customization overrides') and when to use it ('Use when the user says...'). The explicit 'Use when' clause keeps it at 3 rather than the 2-cap for missing trigger guidance.

Provides natural phrasings a user would actually say — 'customize bmad', 'override a skill', 'change agent behavior', 'customize a workflow' — covering common variations. Not below 3 because the triggers are natural rather than jargon.

A clear BMad-specific niche with distinct triggers tied to 'customize bmad'/'override a skill', unlikely to fire for unrelated skills. Not 2 because the triggers are narrow and domain-bound rather than broadly overlapping.