Conciseness

The ~35-line body is lean and every line instructs (rules, fix ladder, workflow, stop conditions); the only explanatory line ("Not a status summary: use `session-compact` for handoff state") is a useful disambiguation, not padding. No explanation of concepts Claude already knows — matches anchor 5.

Actionability

Concrete, executable guidance throughout: exact fix-ladder rungs ("test, lint rule, hook, CI check" / "script, template, default config" / "skill, `AGENTS.md` / `CLAUDE.md`"), specific evidence commands (`git log` and diff, hook/CI failures, `task` evidence), exact record fields, and named tools (`verify`, `session-compact`, `dev-debug`). Per the rubric's code-vs-instruction note, an instruction-only skill with this level of specific guidance earns the top anchor.

Workflow Clarity

Six clearly sequenced steps with explicit validation checkpoints: "Present the items and wait for approval", and "validate each with `verify` (a new check must fail on the old state when practical)" — an explicit feedback criterion. The Stop section adds error-recovery/pause conditions and handoff. Matches anchor 5's clear sequence with explicit validation and error handling.

Progressive Disclosure

The skill is under 50 lines, self-contained with no bundle files, and organized into clear sections (Rules, Fix Ladder, Workflow, Stop). The rubric's simple-skill exception applies: well-organized sections suffice for a 5 when there is no need for external references.

Specificity

The description names the domain and two concrete actions — "Run an evidence-based retrospective on finished work (feature, session, incident) and turn friction into durable fixes, automation first" — with minor gaps (it doesn't enumerate the retro's outputs like the fix ladder or approval workflow). It lists several specific elements without being comprehensive, matching the anchor 4 example rather than anchor 3's 1-2 actions or anchor 5's full coverage.

Completeness

It explicitly answers both 'what' (evidence-based retrospective on finished work; turn friction into durable fixes, automation first) and 'when' (an explicit 'Use when' clause listing concrete trigger phrases). Both are clearly and explicitly stated with concrete triggers, matching anchor 5 rather than anchor 4 where the 'when' would be less specific.

Trigger Term Quality

"Use when users ask for a retro, lessons learned, post-mortem, or 'what went wrong', or when another skill offers one after a feature or fix" comprehensively covers the natural phrases users would say, including synonyms (retro, lessons learned, post-mortem) and a quoted colloquial trigger. This is comprehensive natural-term coverage, matching anchor 5.

Distinctiveness Conflict Risk

The retro/lessons-learned niche is distinct, and the skill even disambiguates itself from neighbors (the body's "Not a status summary: use `session-compact`" and handing unclear failures to `dev-debug`). Minimal conflict risk with a clear niche and distinct triggers, matching anchor 5.