Conciseness

Lean and efficient throughout: a one-line input contract ("Given a single break row (key, GL values, subledger values, bucket, likely cause)..."), terse per-side attribute lists, a statement template, exactly one worked example per break type, and an exact output schema. Nothing explains concepts Claude already knows; every token earns its place.

Actionability

Concrete guidance dominates: exact fields to fetch on each side, exact diff attributes ("posting date, FX rate/date, account mapping, quantity sign, amount sign"), a statement template with four worked examples, and a copy-paste JSON output shape. Not 5 because the fetch steps say "via the internal-gl MCP, fetch..." without naming the actual tools or call syntax, leaving the invocation to inference; not 3 because the guidance is specific and executable far beyond pseudocode level.

Workflow Clarity

The sequence is clear and ordered (pull GL side → pull subledger side → diff attributes → write root-cause statement → emit JSON), and the diff step functions as an implicit checkpoint; the skill is explicitly non-destructive ("this skill diagnoses, it does not post"), so the destructive/batch validation cap does not apply. Not 5 because there is no explicit handling of failure cases — what to do when no matching subledger transaction is found or multiple candidates match.

Progressive Disclosure

The skill is under 50 lines, single-purpose, needs no external references, and is organized into well-signaled sections (Trace path, Cause → statement, Output). Per the rubric's simple-skill guidance, this merits a 5 with no bundle files required.

Specificity

The description lists multiple concrete actions — "Root-cause a reconciliation break to its source transaction or posting", "follow the audit trail from the break row back to the originating entry on each side", and "state what differs and why" — with comprehensive coverage of the skill's scope. No gaps; the description fully reflects what the body delivers.

Completeness

Explicitly answers both questions: the "what" (trace to source, state what differs and why) and the "when" ("Use after gl-recon has classified a break") with a concrete, explicit trigger clause. Matches the anchor-5 example structure of what + explicit when.

Trigger Term Quality

Good natural domain keywords ("reconciliation break", "root-cause", "audit trail", "source transaction or posting", "gl-recon") but a few common variations users might say are missing (e.g. "rec break", "GL break", "subledger", "out of balance"). Not 5 because anchor-level synonym/variation coverage is incomplete; not 3 because the terms present are exactly what a recon user would say.

Distinctiveness Conflict Risk

A clear niche (reconciliation break root-causing) tied to a specific upstream skill and distinct vocabulary (gl-recon, GL, subledger, posting), so it is unlikely to trigger for the wrong skill. Minimal overlap risk with generic skills.