Conciseness

Nearly every section carries non-obvious domain knowledge (PGLite --follow requirement, apply-defaults-to-dry-run gotcha, 72h restore TTL, lock semantics) with no explanation of concepts Claude already knows. Not 5 because of real redundancy: two overlapping anti-pattern sections ("Anti-patterns" and "Anti-Patterns" both warn against running unify-types under autopilot), a decorative ASCII celebration summary, and version-specific details ("v0.41.22", "KNOBS_HASH_VERSION bump (5→6)") that could be trimmed. Not 3 because the waste is confined to a handful of lines in an otherwise lean document.

Actionability

Fully executable throughout: copy-paste commands for every phase (gbrain schema active --json | jq, onboard --check --explain, jobs submit unify-types with the exact --params JSON including the apply:true requirement and the --follow variant for PGLite), a ready SQL rollback statement, gbrain restore/schema use revert commands, and concrete expected outputs for verification. Matches the 5 anchor: specific commands cover the common cases with the key gotchas called out.

Workflow Clarity

A destructive batch migration (25K retypes, 5.5K soft-deletes) with a clearly sequenced 5-phase workflow and explicit validation: Phase 1 checks the active pack and expected identity string, Phase 2 mandates dry-run preview with 'if the proposed changes look wrong, don't proceed', Phase 4 lists exact expected finding transitions (pack_upgrade_available → ok, ≤16 types), plus rollback paths, failure modes with checkpoint resume, and a decision tree. Feedback loops (queued-job → resubmit with --follow, phase failure → resume from checkpoint) are present, matching the 5 anchor; the destructive-operation cap at 3 does not apply because validation is explicit.

Progressive Disclosure

No bundle files exist, so everything is inline in one ~260-line file, but it is well-sectioned with clear headers and logical phase order rather than a wall of text. Not 5 because the body carries a relative reference outside the skill bundle ("../conventions/agent-operator-protocol.md") that is unverifiable from the skill's own directory, and content like the Output Format/Contract detail could plausibly live in a reference file. Not 3 because the structure is good and the three external Reference URLs are clearly signaled one-level-deep links.

Specificity

Lists multiple concrete actions with full specificity: "Migrate a brain from gbrain-base (or any pack) to gbrain-base-v2's 14-canonical-type taxonomy via gbrain onboard --check + the unify-types Minion handler" and "Collapses 94 noisy types to 15 canonical with subtypes, alias rows, and link rows" — named tools, named handler, and quantified outcomes. Not below 4 because coverage of what the skill does is comprehensive, not just 'several specific actions with minor gaps'.

Completeness

Explicitly answers both: what — "Migrate a brain ... Collapses 94 noisy types to 15 canonical with subtypes, alias rows, and link rows"; and when — "Triggers when an agent notices pack_upgrade_available, type_proliferation, or asks 'what is the canonical taxonomy / how do I clean up my page types'". The explicit trigger clause with concrete phrases matches the 5 anchor exactly; the 4 anchor requires a 'when' that 'could be more explicit', which does not apply here.

Trigger Term Quality

Includes natural phrases users would say ("how do I clean up my page types", "what is the canonical taxonomy") plus the system findings that surface the need (pack_upgrade_available, type_proliferation), but misses common variations like "consolidate my types", "too many page types", or "type sprawl" that appear in the skill's own triggers list. Not 5 because synonym coverage is incomplete; not 3 because the terms present go beyond a single generic keyword.

Distinctiveness Conflict Risk

Clear niche with near-zero conflict risk: gbrain pack migration to a named pack version (gbrain-base-v2) via a named PROTECTED handler (unify-types) — no other plausible skill would trigger on these terms. Matches the 5 anchor ('clear niche with distinct triggers; minimal conflict risk').