Conciseness

The body is dense and load-bearing — exact statuses, commands, and report formats with no basic-concept padding — but the <context> block ("A session usually ends when the agent says 'done'. That sentence is a claim, and so is a ticked box in a task file") is rationale Claude could partly infer and could be trimmed. Not 3: there are no unnecessary library or format explanations; not 5: the context paragraph and a few explanatory sub-clauses are mild over-explanation.

Actionability

Fully actionable instruction-only guidance: exact status definitions ("met", "not met", "not verified", "withdrawn"), exact commands ("git status --porcelain", "git rev-list --count @{u}..HEAD"), and copy-paste verdict templates ("Goals: all met (<N> of <N>)"). Not 4: the commands and output templates are complete and executable with no gaps, which the scoring notes state is not penalized for lacking code.

Workflow Clarity

Six clearly sequenced steps with an explicit verification checklist, edge-case handling ("A ticked checkbox whose evidence command fails is not met"), a defined empty-input path ("No goal found... skip to the last step"), and an explicit TOTAL STOP endpoint. Not 4: validation checkpoints and error-recovery paths are explicit rather than implicit. The skill is read-only, so the destructive/batch validation cap does not apply.

Progressive Disclosure

A self-contained single-procedure skill with no external references needed, organized into clear sections (Overview, Context, Rules & Constraints, Instructions, Verification). Not 4: nothing inlined belongs in a separate file and navigation is trivial; per the simple-skill note, well-organized sections with no need for external references warrant the top score.

Specificity

The description enumerates concrete actions covering the entire workflow — "lists everything they asked for, gathers evidence for each item and reports met or not met, changing nothing" — with no coverage gaps. Anchor 4 would require minor gaps in coverage, which are absent.

Completeness

Both halves are explicit: the "what" ("lists everything they asked for, gathers evidence for each item and reports met or not met, changing nothing") and the "when" ("Use when the user asks whether the goals of the CURRENT session are met") with a concrete trigger phrase. This matches the anchor-5 example structure exactly, and a "Use when..." clause is present so no cap applies.

Trigger Term Quality

"Use when the user asks whether the goals of the CURRENT session are met" is a natural phrase a user would say, and "reflect" anchors the negative boundary. It falls short of 5 because common synonyms like "done", "complete", or "finish" are missing, but exceeds 3 since the core trigger phrasing is direct rather than generic.

Distinctiveness Conflict Risk

"Not for auditing how the session went — use reflect" explicitly disambiguates from the closest sibling skill, and the trigger (session goals being met) is a clear niche. Not 4: the explicit negative boundary plus specific trigger leaves minimal overlap risk with other skills.