Adversarial Self-Audit Prompts
A seven-prompt pattern that turns a chat model into an adversary against your own business or plan, surfacing weak spots before the market does. Marketed as "Claude Red Team Mode," but it is a prompt technique, not a product feature. Source: X/@free_ai_guides thread, 2026-05-27
The premise
Most people use a model to validate ideas. This pattern does the opposite: it asks the model to attack the plan from the perspectives most likely to break it. Run all seven for a full adversarial audit, or pick the single lens that targets the current biggest blind spot. Source: X/@free_ai_guides thread, 2026-05-27
The seven lenses
- Competitor War Game: the model role-plays a ruthless competitor and builds the exact strategy to steal customers, undercut positioning, or make the product irrelevant within 12 months, with named moves and a timeline.
- Customer Betrayal Sim: a churned customer writes an honest breakup email explaining what triggered the cancellation and where they went instead.
- Investor Teardown: a skeptical VC names the three biggest holes, the partner-meeting question you cannot yet answer, and the comparable company that tried this and failed.
- Single Point of Failure Scan: audits people, tools and platforms, revenue, and supply chain for dependencies, rating each by likelihood and severity.
- Market Shift Scenario: generates three plausible 18-month shifts (regulatory, technological, behavioral), their impact, early warning signs, and the required pivot.
- Pricing Pressure Test: a price-sensitive buyer argues the product is not worth its price, then identifies the exact churn price point and the competitor that makes your pricing look bad.
- Attack Surface Report: compiles a structured red-team briefing with the top three vulnerabilities (ranked), the single biggest blind spot, and a hardened version of the plan.
Source: X/@free_ai_guides thread, 2026-05-27
Relevance to Kevin
This is the business-strategy analogue of the security red-teaming Kevin already runs in code: Security and Review Skills attacks a repo, Frontend and Design Skills enumerates abuse paths, and this pattern attacks a business model. It pairs naturally with the Services-as-Software filter as a pre-mortem: before committing to a venture, stress-test the moat, pricing, and single points of failure adversarially rather than seeking confirmation.
Executable route: use the red-team-business skill when Kevin asks to attack a business, product, launch, pricing decision, or project plan. The skill turns the seven lenses into a ranked vulnerability report with falsification checks instead of leaving the prompts as inspiration. Source: X bookmark artifact audit, 2026-06-30
Concept Position
| Field | Value |
|---|---|
| Concept family | Agent harness and runtime primitives |
| Concept owned | A seven-prompt pattern that turns a chat model into an adversary against your own business or plan, surfacing weak spots before the market... |
| Category map | Concept System Map |
Timeline
- 2026-07-01 | Concepts category refresh added this page to the Agent harness and runtime primitives family, linked it to Concept System Map, and kept it standalone because it owns this reusable mental model: A seven-prompt pattern that turns a chat model into an adversary against your own business or plan, surfacing weak spots before the market... Source: User request, 2026-07-01
- 2026-05-27 | @free_ai_guides posted the seven-prompt "Red Team Mode" thread (2,665 likes, 4,918 bookmarks). The eighth and ninth replies were a promo for a paid Claude skills bundle, not part of the framework. Source: X/@free_ai_guides, 2026-05-27
- 2026-06-30 | Deep bookmark review added the executable
red-team-businessskill and kept the Linktree promo replies out of the framework. Source: X bookmark artifact audit, 2026-06-30