insight-adversarial

Runs multi-pass adversarial critique of plans and designs using dynamically selected critic personas.

14|1|Updated Jun 14, 2026
One-click install
npx skills add https://github.com/allemaar/open-skills --skill insight-adversarial-allemaar
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: insight-adversarial
Source: https://github.com/allemaar/open-skills/tree/main/skills/insight-adversarial
Command: npx skills add https://github.com/allemaar/open-skills --skill insight-adversarial-allemaar

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Plans, designs, documents, and ideas often carry hidden assumptions and blind spots that a single review pass misses. This Skill subjects a target to structured adversarial scrutiny from multiple distinct critic personas, surfacing critical issues before they become costly failures. ## Core Features & Use Cases - Multi-Pass Persona Critique: Runs 1–3 passes, each selecting 4+ fresh personas from a roster of 20+ critics (Skeptic, Security Auditor, CFO, Red Team Lead, and more), with no persona reused across passes. - Target Classification: Classifies the target (plan, design, YON workflow, document, idea, business case, code, or UI/UX) to select the most relevant personas for that failure mode. - Structured Synthesis: Produces ranked critical issues, aggregate blind spots, minimum surgical remediations, and a one-line verdict after each pass. - Use Case: Before executing a project plan, ask for an adversarial review; the Skill attacks it from the Skeptic, Pragmatist, Scope Cop, and Burned Engineer angles, then delivers a verdict on whether to proceed or redesign. ## Quick Start Ask the agent to red team your plan or stress-test your design document using the insight-adversarial skill.

Frequently Asked Questions about insight-adversarial

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I stress-test a plan before executing it?

Invoke the skill with a request like "red team this plan" or "attack my plan". It classifies the target, selects 4–5 relevant critic personas, runs each through a focused critique pass, and delivers a synthesized verdict with ranked critical issues.

What is multi-persona adversarial review?

It is a structured critique method where distinct critic personas (Skeptic, Security Auditor, CFO, Competitor, and others) each attack the same target from their own lens. Findings are synthesized into critical issues, blind spots, and remediations rather than a single opinion.

How many review passes does the adversarial critique run?

It runs 1–3 passes. Pass 1 uses 4–5 personas, Pass 2 adds 2–4 new personas focused on unresolved issues, and Pass 3 delivers a final verdict. Personas are never reused across passes.

What types of targets can be adversarially reviewed?

The skill supports eight target types: plans, designs/architecture, YON workflows, documents/specs, ideas/concepts, business/product cases, code/implementations, and UI/UX. Persona selection is tailored to each target type's primary failure mode.

When should I not use adversarial multi-pass critique?

Avoid it for trivial or low-stakes artifacts where a single review suffices, since each pass requires multiple persona critiques and user confirmation between passes. It is designed for targets where missed assumptions carry real cost.