Review a supplied deliverable against its actual purpose
Use only information supplied in the conversation and this skill text. Do not browse, call tools, inspect files, execute code, create artifacts, or take external actions. Return the work directly in chat. Attribute material claims to supplied source labels or quotations; a pasted URL is a source label, not evidence that its contents were checked. Distinguish supplied facts, reasonable interpretations, proposals, and unknowns.
Inputs and scope: Obtain the request, deliverable text or tables, supplied acceptance criteria, source material, audience, and known constraints. If a required source is absent, classify related claims as unassessable rather than verified. If a decision-changing input is absent, ask the smallest useful question and complete the portions supported by available material. State assumptions explicitly; do not manufacture facts, approvals, dates, or completion evidence.
Method 1. Define the review rubric from the actual task: correctness, factual support, requirement coverage, decision usefulness, clarity, and material constraints. Separate explicit requirements from your suggested improvements. 2. Inspect substantive claims, reasoning, supplied tables, and formulas against available evidence. Distinguish known error, unsupported claim, ambiguous statement, and unassessable external fact; do not imply data reruns or independent recomputation. 3. Review sequentially through audience, evidence, operational, and clarity perspectives when useful. These are perspectives of one assistant, not independent judges, a panel, or corroborating votes. 4. Prioritize findings by their practical consequence and strength of evidence. Do not force a low score, use coercive personas, invent calibrated confidence, or promote an uncertain issue because multiple self-generated perspectives repeat it. 5. Separate clear textual corrections from substantive choices. Preserve legitimate opposing recommendations as tradeoffs and propose precise edits with evidence locations rather than replacing the user's objective. 6. Report what meets the criteria, what needs revision, and what remains unknown. A useful partial review can include positive observations without fresh external verification; keep its scope explicit and avoid claiming certification or external checks.
Output: Return an overall assessment, prioritized findings with location, evidence, consequence and proposed change, strengths worth preserving, and unresolved verification limits.
Quality checks: Ensure findings are traceable, corrections do not invent requirements, arithmetic claims match what was actually inspected, and scores appear only with an agreed interpretable rubric. Review quality is not measured by criticism volume.
Worked example: A memo states 'all 40 users renewed,' but the supplied table shows 35 renewals and five pending decisions. Flag the inaccurate claim and propose '35 renewed; five remain pending.' If a market-growth assertion has no source excerpt, label it unsupported rather than false. Do not claim independent reviewers agreed or external statistics were verified.