Conciseness

The ~44-line body is lean and assumes competence throughout: "Listing the callers is not the job. The agent can grep those in a second" and "Work out when things run: microtasks, unmount and teardown, Solid versus React" pack craft-specific guidance with zero padding. It never explains concepts Claude already knows, so it matches anchor 5 ("every token earns its place"); the deliberate repetition of "the one fact it's safe because of" is a structural through-line, not over-explanation, so it does not drop to 4.

Actionability

As an instruction-only skill it gives concrete, executable direction: the 5-level evidence ladder ("You ran it. A script or test that calls the real code and fails loud if you're wrong"), "check its pinned version and any local patch", and "Write a script or test that runs the real code, run it, and paste what happened". It scores 4 rather than 5 because a worked example is absent — there is no sample proof script or example of the pasted output and what a hand-back actually looks like, which anchor 5 ("specific examples cover the common cases") would require; it is well above anchor 3 because the guidance is real steps, not pseudocode or high-level hints.

Workflow Clarity

Six clearly sequenced steps with an explicit verification protocol woven in: the evidence ladder requires the agent to "say where it stopped", step 5 says to run the proof and "paste what happened", and the hand-back checklist demands "show the proof. If you couldn't prove it, write unproven" — a real error-recovery fallback. This matches anchor 5 (clear sequence, explicit validation, checklist); the destructive/batch cap does not apply since this is a read/verify workflow.

Progressive Disclosure

The body is under 50 lines with no bundle files (no references/, scripts/, or assets/ exist) and no content that needs offloading — well-organized sections (Steps, How sure are you, What to hand back) with clear headers. Per the rubric's simple-skill guidance this earns a 5; references to sibling skills like `why`, `how`, `unslop`, and `arena` are cross-skill pointers, not buried file references, and there is no nested-reference structure to penalize.

Specificity

"Find what a change could break somewhere else before it ships, beyond the diff" and "prove the one fact it's safe because of by running real code" name the domain plus several concrete, vivid actions (find downstream breakage, prove the safety fact, run real code rather than write up). It falls between anchor 3 ("1-2 concrete actions") and anchor 4 ("several specific actions") but is noticeably above the midpoint because the actions are sharply concrete rather than generic; it is not a 5 since coverage of the full workflow (reading library source, checking pinned versions, citing file:line) lives only in the body.

Completeness

It clearly answers both questions: the 'what' ("Find what a change could break somewhere else before it ships... and prove the one fact it's safe because of by running real code") and the 'when' with concrete trigger phrases ("Use for 'blast radius of X', 'what could this break', or reviewing a small diff you don't trust"), matching the anchor-5 example structure exactly. It is not below 5 because neither half is vague or merely implied.

Trigger Term Quality

"Use for 'blast radius of X', 'what could this break', or reviewing a small diff you don't trust" gives good natural-phrase coverage that users would plausibly say verbatim. It sits between anchor 4 ("good keyword coverage; a few natural terms missing") and anchor 5 because it omits common synonyms and variations like "impact analysis", "what does this affect", or "downstream/upstream dependencies".

Distinctiveness Conflict Risk

"blast radius" and "what could this break" are distinct triggers carving a clear niche, but "reviewing a small diff you don't trust" overlaps with general code-review skills, giving minor overlap risk with a closely related skill (anchor 4) rather than the minimal conflict risk of anchor 5.