Conciseness

The body is lean and assumes competence — the pre-mortem vs red-team distinction and the 'Five real kill-assumptions beat twenty generic risks' framing add genuine signal, and the 'Fails if' section even models bad-vs-good phrasing ('"execution risk"' example). Minor over-explanation: the Notes section restates instructions 2, 4, and 5 ('No strawmanning', 'No fabrication', 'Rank ruthlessly'), which could be trimmed. Efficient overall with a few trims available — the 4 anchor, not 5, because some redundancy exists.

Actionability

For an instruction-only skill the guidance is fully executable: a numbered 8-step process, a concrete falsifiability pattern ('Write each failure mode as "Fails if ___."') with an explicit good/bad comparison, a ranking formula, and a copy-paste output template specifying every field per assumption. Specific examples cover the common cases — matches the 5 anchor for instruction skills (concrete, specific guidance with nothing vague).

Workflow Clarity

Steps are clearly sequenced (extract claims → steelman → attack → write failure modes → rank → self-refute → cross-model option → structure output) and the output template doubles as a final checklist. No explicit validation/feedback loops, so it does not reach the 5 anchor, but this is a read-only analytical skill rather than a destructive/batch operation, so the 3-cap for missing validation does not apply — the 4 anchor (clear sequence, most checkpoints present) is the best fit.

Progressive Disclosure

The skill has no bundle files, needs none, and everything lives in one well-organized file with clear sections (Purpose, Context, Instructions, Notes, Further Reading). External links are one level deep and clearly labeled under 'Further Reading'. No nesting, no monolithic overload, no content that belongs in a separate file — matches the well-organized single-file case for a self-contained skill; under the simple-skill guidance this earns full marks with just well-organized sections.

Specificity

The description lists multiple concrete actions covering the full workflow — 'Steelmans then attacks each claim, ranks failure modes by impact × likelihood × cheapness-to-test, and returns the cheapest test and kill criteria for each' — all in third person with no vague language. This matches the 5 anchor (multiple specific concrete actions, comprehensive coverage); nothing is generic or padded.

Completeness

Explicitly answers both: 'what' (red-team a PRD/roadmap/strategy by attacking load-bearing assumptions, steelman then attack, rank, return tests and kill criteria) and an explicit 'Use when...' clause with four concrete trigger phrases. This is a direct match for the 5 anchor; not 4 because the 'when' is not merely present but specific and multi-scenario.

Trigger Term Quality

Good natural keyword coverage: 'stress-testing a plan, pressure-testing a strategy, challenging assumptions, or preparing a doc for executive review' plus 'PRD, roadmap, or strategy'. A few natural synonyms users might say are missing (e.g., 'red team', 'pre-mortem', 'poke holes in'), so it sits between the 4 anchor (good coverage, a few natural terms missing) and 5 — not 3 since coverage is clearly more than partial.

Distinctiveness Conflict Risk

The red-team niche (PRD, roadmap, strategy stress-testing) is mostly distinct from general review/critique skills, with dedicated triggers like 'stress-testing a plan'. Minor overlap risk remains: 'challenging assumptions' and 'preparing a doc for executive review' could plausibly invoke general document-review or pre-mortem skills, keeping it just below the 'minimal conflict risk' of a 5.