# Cold Offer Review > Audit an existing B2B offer from the perspective of a cold buyer with no transferred trust. Use when reviewing a website, proposal, call transcript, case study, offer note, pricing or terms, or user-supplied answers to determine what is clear, supported, coherent, and safe to test; expose contradictions, claim-to-proof gaps, risk, effort, time-to-value, and first-commitment weaknesses without claiming market validation or likely conversion. Do not use to invent, redesign, or build an offer; route an explicitly requested post-audit rebuild to offer-creation. - Skill: `moussaoui-ghiles/cold-offer-review` (Agent Skill, multi-file: 5 files) - Install (CLI): `npx skillmds@latest add moussaoui-ghiles/cold-offer-review` - Raw SKILL.md: https://api.skillmd.com/api/skills/moussaoui-ghiles/cold-offer-review/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Security - Author: Moussaoui-Ghiles (https://skillmd.com/u/moussaoui-ghiles) - Updated: 2026-09-21 - Page: https://skillmd.com/skills/moussaoui-ghiles/cold-offer-review --- # Cold Offer Review Audit whether a cold buyer can understand, believe, judge, and accept an existing B2B offer. Treat the result as an evidence-led diagnosis, never as market validation or a conversion forecast. ## Required references Read all three before auditing: - [Intake and evidence protocol](references/intake.md) for workspace search, progressive questions, evidence labels, and input coverage. - [Audit criteria](references/audit-criteria.md) for the seven dimensions, additional cold-buyer tests, status rules, source limits, and earliest-broken-link logic. - [Output schema](references/output-schema.md) for the required report and validation plan. ## Scope boundary - Review an existing offer. Accept websites, proposals, call transcripts, case studies, notes, offer and pricing documents, terms, and user answers. - Audit the offer as presented to the stated buyer and sales context. Do not silently rewrite it. - Do not declare market urgency, product-market fit, offer validation, or likely conversion from copy or internal coherence alone. - Do not invent buyer pains, proof, testimonials, guarantees, scarcity, outcomes, prices, or terms. - Do not import universal benchmarks. Treat external frameworks and operator examples as diagnostic prompts, not laws or proof for this offer. - Route net-new offer creation or a requested redesign/rebuild to `$offer-creation`. When the original request is "audit, then improve," finish the evidence-led audit first, then invoke `$offer-creation` in the same turn using only the evidenced findings. Ask for further authority only if the rebuild requires a new material business decision. ## Workflow ### 1. Fix the audit question State: - offer and version or date; - intended buyer; - true cold, trust-assisted, warm follow-up, mixed, or unknown sales context; - artifacts in scope; - decision the audit should inform. Mark any missing item `unknown`. Never substitute a likely answer. ### 2. Search before asking Search the user-provided workspace for the company, offer name, buyer, mechanism, pricing, and evidence terms before asking questions. Inspect only relevant offer notes, proposals, transcripts, case studies, terms, and decision records. Prefer exact-name searches, then bounded keyword searches. Treat private notes as input, not automatically as buyer-visible proof. Record inaccessible links or files as unknown. Do not widen into unrelated confidential material. ### 3. Build the evidence register Extract claims and inputs without merging their status. Label every item as one of: - `current decision`; - `verified fact`; - `user-provided assertion`; - `source/operator example`; - `inference`; - `unknown`. Record the artifact, location, date, buyer visibility, and limitation. Never turn missing values into zero, false, or a fact. `Current decision` means the user or an authoritative decision record identifies the rule or version that governs this audit. It establishes decision authority, not the factual truth of a buyer, market, performance, or legal claim. ### 4. Complete intake progressively Use the intake reference. Populate it from artifacts first. Ask only the smallest next batch of high-leverage questions, normally three to five, and explain which audit decision each unlocks. Continue in batches when needed. If the offer nucleus is incomplete or artifacts conflict, return the supported portion of the audit, preserve each contradiction or gap, and ask the smallest governing question. Do not manufacture a pass/fail verdict to fill the template. ### 5. Map claims to proof Create a separate row for each material promise, numerical claim, guarantee premise, testimonial implication, timing claim, economic claim, and differentiator. Match only evidence that supports the same buyer, problem, method, result, conditions, period, acquisition motion, buyer visibility, and permission at the claimed scope. An artifact can verify that a claim was made without verifying that the claimed result occurred. State that distinction. ### 6. Audit the offer Apply every required dimension and additional test in the criteria reference. Label every conclusion `PASS`, `FAIL`, or `UNPROVEN` and cite the supporting artifact or missing evidence. - Use `FAIL` only for an evidenced contradiction, incoherence, or explicit mismatch. - Use `UNPROVEN` when evidence is missing, inaccessible, stale for the claim, or insufficiently matched. - Use `PASS` only for the bounded question tested; a pass never means the market will buy. Expose contradictions instead of resolving them by assumption. ### 7. Diagnose the earliest broken link Trace the cold-buyer decision chain in order. Report the earliest evidenced failure and the earliest unresolved link separately. If no failure is proven, say so; if no link is unresolved, say so. Never call uncertainty a break. Prioritize that link over cosmetic copy improvements. ### 8. Define the next evidence and validation step Specify the exact artifact, observation, buyer response, operational record, or term clarification needed next. Propose the smallest safe test that can resolve the highest-impact uncertainty. Predefine what the test will inform. Do not promise validation from an arbitrary number of interviews, replies, or sales calls, and do not borrow an operator's benchmark as a decision threshold. ### 9. Deliver the report Use the output mode in the schema. Default to the compact decision brief. Use the full audit only when the user asks for it or when the scope, stakes, or contradictions require the underlying tables. Lead with the status and earliest broken or unresolved link. ## Hard-stop conditions Stop only when no bounded audit is possible because: - the buyer or offer itself is not identifiable; - an essential artifact is inaccessible and no supplied evidence supports even a partial review; - the only requested conclusion is legal and requires qualified legal review; or - the requested conclusion cannot be narrowed to the available evidence. Unknown sales context, conflicting price or scope, incomplete numerical claims, and unspecified guarantee obligations do not stop the whole audit. Mark affected conclusions `UNPROVEN` or `FAIL` under the criteria, complete the supported review, and ask what governs.