返回 Skill 列表
extension
分类: 开发与工程无需 API Key

red-blue-validator

通过红/蓝队动态进行迭代对抗压力测试。红队对提案发起实质性的、强有力的攻击;蓝队则以防御、缓解措施和强化来回应。这一过程将持续循环,直到满足收敛标准,从而产生经过实战检验的提案。主动激活用于:(1) 需要进行压力测试的重要决策,(2) 在重大承诺前验证策略,(3) 加固架构决策,(4) 准备提案辩护,(5) 审查安全态势,(6) 从对手的角度进行投资尽职调查。触发词包括:“红队这个”,“蓝队”,“压力测试”,“攻击这个计划”,“找出弱点”,“对抗审查”,“唱反调者”,“可能会出什么问题”,“挑刺”,“挑战这个决定”,“模拟战争”。

person作者: jakexiaohubgithub

Red/Blue Team Validator

"Find weaknesses before reality does."

Every proposition—whether a decision, strategy, architecture, or plan—has vulnerabilities. This skill systematically exposes them through iterative adversarial cycles. Red Team attacks with substantive, steel-manned challenges. Blue Team defends with mitigations and hardening. The cycle continues until convergence: a battle-tested proposition with documented defenses.


1. Purpose

Core Value Proposition

Static analysis misses what adversarial pressure reveals. Red/Blue validation simulates the attacks your proposition will face—from competitors, critics, reality itself—and forces you to build defenses before you need them. The output is not just a risk list, but a hardened proposition that has survived systematic assault.

Capabilities

| # | Capability | Phase | Value | |---|-----------|-------|-------| | 1 | Proposition intake with attack surface mapping | Pre-Round | Define what can be attacked | | 2 | Experience pool loading (domain failure patterns) | Pre-Round | Avoid reinventing known failures | | 3 | Multi-category attack generation | Round N: Red | Surface vulnerabilities systematically | | 4 | Steel-manning attacks to maximum strength | Round N: Red | Ensure attacks are not strawmen | | 5 | Severity scoring (CRITICAL/HIGH/MEDIUM/LOW) | Round N: Red | Prioritize responses | | 6 | Defense generation (REFUTE/MITIGATE/ACCEPT/HARDEN) | Round N: Blue | Address each attack | | 7 | Proposition hardening through iterative refinement | Round N: Blue | Strengthen against attacks | | 8 | Convergence evaluation with explicit criteria | Round N: Eval | Know when to stop | | 9 | RISK-ASSESSMENT synthesis (CONTRACT-08) | Post-Round | Standardized output | | 10 | Hardened proposition generation | Post-Round | Battle-tested version | | 11 | Attack/defense log compilation | Post-Round | Audit trail | | 12 | Go/no-go recommendation | Post-Round | Decision support |


2. When to Use

Ideal Use Cases

| Scenario | Why Red/Blue Validation Matters | |----------|--------------------------------| | Pre-commitment decision review | Simulate objections before committing resources | | Strategy validation | War-game competitive responses and market realities | | Architecture decision hardening | Stress-test technical choices before implementation | | Proposal defense preparation | Anticipate and prepare for stakeholder pushback | | Investment due diligence | Adversarial review of financial projections and market assumptions | | Security posture assessment | Systematic attack surface enumeration | | Go/no-go decisions | High-stakes decisions need adversarial pressure | | Policy/process validation | Find edge cases and failure modes | | Product launch readiness | Anticipate market, competitive, and operational challenges | | M&A target evaluation | Adversarial review of synergy claims |

Anti-Patterns (When NOT to Use)

| Anti-Pattern | Why It's Ineffective | Better Alternative | |--------------|----------------------|-------------------| | Low-stakes decisions | Over-engineering for trivial choices | Just decide and iterate | | Time-critical emergencies | Fires need extinguishing, not philosophy | Act, then debrief | | Already committed | Adversarial review after commitment creates conflict | Use for future decisions | | Early exploration | Premature to attack ideas still forming | Use after initial validation | | Confirmation theater | Going through motions without genuine adversarial intent | Either commit to true adversarial thinking or skip | | Reversible decisions | Two-way doors don't need siege testing | Save intensity for one-way doors |


3. Parameters

| Parameter | Type | Required | Default | Description | |-----------|------|----------|---------|-------------| | subject_type | enum | yes | — | decision | strategy | architecture | plan | policy | investment | security | | max_rounds | integer | no | 3 | Maximum red/blue cycles (1-5) | | attack_intensity | enum | no | standard | light | standard | aggressive | | attack_categories | list | no | auto | Categories to probe (see catalog); auto selects by subject_type | | convergence_mode | enum | no | no_new_critical | no_new_critical | all_addressed | round_limit | | include_experience_pool | boolean | no | true | Load domain-specific failure patterns | | steel_manning_level | enum | no | standard | minimal | standard | maximum | | output_mode | enum | no | full | risk_assessment | hardened_proposition | full_log |

Parameter Effects Matrix

| Parameter | Red Phase Effect | Blue Phase Effect | Convergence Effect | |-----------|-----------------|------------------|-------------------| | attack_intensity: light | Top 3 attack categories | Quick defenses | max_rounds capped at 2 | | attack_intensity: standard | Top 5 attack categories | Full defense protocol | Normal convergence | | attack_intensity: aggressive | All applicable categories | Exhaustive defense | Requires no_new_critical | | steel_manning_level: minimal | 1-pass attacks | — | Faster rounds | | steel_manning_level: standard | 2-pass steel-manning | — | Normal rounds | | steel_manning_level: maximum | 3-pass with ideological Turing test | — | Thorough rounds | | convergence_mode: no_new_critical | — | — | Stop when 0 new CRITICAL/HIGH | | convergence_mode: all_addressed | — | Must address all | Stop when no ACCEPT responses | | convergence_mode: round_limit | — | — | Stop at max_rounds |

Auto-Selected Attack Categories by Subject Type

| Subject Type | Default Attack Categories | |--------------|--------------------------| | decision | ASSUMPTIONS, ALTERNATIVES, REVERSIBILITY, CONSEQUENCES, TIMING | | strategy | COMPETITIVE, MARKET, EXECUTION, DEPENDENCIES, TIMELINE | | architecture | SCALABILITY, SECURITY, DEPENDENCIES, OPERATIONAL, EDGE_CASES | | plan | FEASIBILITY, RESOURCES, TIMELINE, DEPENDENCIES, RISKS | | policy | EDGE_CASES, ENFORCEMENT, UNINTENDED_CONSEQUENCES, POLITICAL | | investment | ECONOMIC, MARKET, EXECUTION, COMPETITIVE, ASSUMPTIONS | | security | ATTACK_SURFACE, VULNERABILITIES, DEPENDENCIES, OPERATIONAL |


4. Checkpoints

This skill uses interactive checkpoints (see references/checkpoints.yaml) to resolve ambiguity:

  • subject_type_classification — When proposition type is ambiguous
  • attack_intensity_selection — When attack intensity not specified
  • convergence_mode_selection — When convergence criteria not specified
  • premature_convergence_check — When convergence met but warning signs present
  • infinite_loop_risk — When defenses generate more attacks than they resolve
  • output_mode_selection — When output format not specified

5. Iterative Workflow

Workflow Overview

┌─────────────────────────────────────────────────────────────────────────────┐
│                       RED/BLUE TEAM VALIDATOR                               │
├─────────────────────────────────────────────────────────────────────────────┤
│                                                                              │
│  ╔══════════════════════════════════════════════════════════════════════╗   │
│  ║                      PRE-ROUND SETUP                                  ║   │
│  ║  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐                   ║   │
│  ║  │ Proposition │  │   Attack    │  │ Experience  │                   ║   │
│  ║  │   Intake    │─▶│  Surface    │─▶│    Pool     │                   ║   │
│  ║  │             │  │  Mapping    │  │   Loading   │                   ║   │
│  ║  └─────────────┘  └─────────────┘  └─────────────┘                   ║   │
│  ╚══════════════════════════════════════════════════════════════════════╝   │
│                                  │                                           │
│                                  ▼                                           │
│  ╔══════════════════════════════════════════════════════════════════════╗   │
│  ║                         ROUND N                                       ║   │
│  ║  ┌─────────────┐                    ┌─────────────┐                  ║   │
│  ║  │  RED TEAM   │                    │  BLUE TEAM  │                  ║   │
│  ║  │   ATTACK    │───── Attacks ─────▶│   DEFENSE   │                  ║   │
│  ║  │ (Generate & │                    │ (Respond &  │                  ║   │
│  ║  │ Steel-man)  │                    │  Harden)    │                  ║   │
│  ║  └─────────────┘                    └─────────────┘                  ║   │
│  ║         │                                  │                          ║   │
│  ║         └────────────┬─────────────────────┘                          ║   │
│  ║                      ▼                                                ║   │
│  ║              ┌─────────────┐                                          ║   │
│  ║              │ EVALUATION  │                                          ║   │
│  ║              │ & Converge? │                                          ║   │
│  ║              └─────────────┘                                          ║   │
│  ║                      │                                                ║   │
│  ║           ┌──────────┴──────────┐                                     ║   │
│  ║           ▼                     ▼                                     ║   │
│  ║    [NOT CONVERGED]        [CONVERGED]                                 ║   │
│  ║    → Round N+1            → Exit loop                                 ║   │
│  ╚══════════════════════════════════════════════════════════════════════╝   │
│                                  │                                           │
│                                  ▼                                           │
│  ╔══════════════════════════════════════════════════════════════════════╗   │
│  ║                     POST-ROUND SYNTHESIS                              ║   │
│  ║  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐                   ║   │
│  ║  │    RISK     │  │  Hardened   │  │  Attack/    │                   ║   │
│  ║  │ ASSESSMENT  │  │Proposition  │  │ Defense Log │                   ║   │
│  ║  │(CONTRACT-08)│  │  Output     │  │             │                   ║   │
│  ║  └─────────────┘  └─────────────┘  └─────────────┘                   ║   │
│  ╚══════════════════════════════════════════════════════════════════════╝   │
│                                                                              │
└─────────────────────────────────────────────────────────────────────────────┘

Pre-Round Setup

Purpose: Prepare the battlefield—understand what's being tested and load relevant knowledge.

Steps:

  1. Proposition Intake

    • Receive subject (decision, strategy, architecture, plan, etc.)
    • If verbal, request written summary or create one together
    • Extract key claims and assertions to be defended
    • Identify stakeholders and constraints
    • Note: Proposition should be specific enough to attack
  2. Attack Surface Mapping

    • Identify dimensions available for attack (from attack-vector-catalog)
    • Map proposition claims to attackable surfaces
    • Select attack categories based on subject_type or explicit attack_categories
    • See: references/attack-vector-catalog.md for categories
  3. Experience Pool Loading (if include_experience_pool: true)

    • Load domain-specific failure patterns
    • Reference historical failures in similar contexts
    • Prepare anti-patterns to probe
    • See: references/experience-pool-patterns.md for patterns
  4. Set Parameters

    • Confirm attack intensity, convergence mode, steel-manning level
    • Estimate expected rounds based on complexity

    CHECKPOINT: subject_type_classification

    • If subject_type not specified or ambiguous: AskUserQuestion
    • Present subject type options with attack category implications

    CHECKPOINT: attack_intensity_selection

    • If attack_intensity not specified: AskUserQuestion
    • Present intensity options with effort implications

    CHECKPOINT: convergence_mode_selection

    • If convergence_mode not specified: AskUserQuestion
    • Present convergence options with trade-offs

Quality Gate: Attack Surface Mapped

  • [ ] Proposition boundaries explicitly defined
  • [ ] Attack categories selected (minimum 3)
  • [ ] Experience pool loaded (if enabled)
  • [ ] Parameters confirmed

Output: Attack-ready proposition with mapped attack surface


Round N: Red Team Phase

Purpose: Generate substantive, steel-manned attacks on the proposition.

Reference: See references/red-team-techniques.md and references/steel-manning-protocol.md.

Steps:

  1. Attack Generation

    For each attack category in scope, generate attacks:

    | Technique | When to Use | Expected Yield | |-----------|-------------|----------------| | Pre-mortem | Always | 3-5 attacks | | Inversion | Strategy, Decision | 2-4 attacks | | Competitor Simulation | Strategy, Investment | 2-3 attacks | | Stress Test Amplification | Architecture, Plan | 2-4 attacks | | Devil's Advocate | Policy, Decision | 2-3 attacks | | Blind Spot Hunter | All | 1-3 attacks | | Historical Pattern Matching | All (with experience pool) | 2-4 attacks | | Black Hat Thinking | Security, Competitive | 3-5 attacks |

    • See: references/red-team-techniques.md for protocols
  2. Steel-Manning (per steel_manning_level)

    For each attack, strengthen to maximum potency:

    | Level | Passes | Protocol | |-------|--------|----------| | minimal | 1 | Basic attack formulation | | standard | 2 | + "How can this be more damaging?" | | maximum | 3 | + Ideological Turing test: "Would a true opponent accept this?" |

    Steel-manning checklist:

    • [ ] Attack is specific, not vague

    • [ ] Attack has clear mechanism of harm

    • [ ] Attack includes realistic trigger conditions

    • [ ] Attack would concern a reasonable proponent

    • [ ] Attack is not easily dismissed

    • See: references/steel-manning-protocol.md for full protocol

  3. Severity Scoring

    Score each attack using SEVERITY-SCORING (RUBRIC-07):

    | Severity | Definition | Response Urgency | |----------|------------|------------------| | CRITICAL | Blocks primary objective; cannot proceed | Must address in Blue Phase | | HIGH | Significant impact; major rework required | Should address in Blue Phase | | MEDIUM | Degrades quality; should fix but can proceed | Address if time permits | | LOW | Minor issue; cosmetic | Document and monitor |

    Scoring dimensions:

    • Impact (0.5 weight): How damaging if attack succeeds?
    • Likelihood (0.3 weight): How likely is this attack vector?
    • Detectability (0.2 weight): How hard to see this coming?
  4. Attack Documentation

    For each attack:

    Attack ID: ATK-[round]-[number]
    Category: [From attack-vector-catalog]
    Target: [What aspect of proposition]
    Statement: [Clear attack formulation]
    Mechanism: [How this would cause harm]
    Severity: [CRITICAL | HIGH | MEDIUM | LOW]
    Steel-manning: [minimal | standard | maximum] - [notes]
    Experience pool match: [Pattern ID if applicable]
    

Quality Gate: Attacks Substantive

  • [ ] Minimum 3 attacks generated
  • [ ] At least 2 different attack categories represented
  • [ ] Steel-manning applied per level
  • [ ] No trivial or easily dismissed attacks
  • [ ] Severities assigned with rationale

Output: Prioritized attack list for Blue Team


Round N: Blue Team Phase

Purpose: Respond to each attack with defenses, mitigations, or proposition hardening.

Reference: See references/blue-team-techniques.md for defense protocols.

Steps:

  1. Attack Triage

    Prioritize attacks by severity:

    • CRITICAL: Must address this round
    • HIGH: Should address this round
    • MEDIUM: Address if time/capacity permits
    • LOW: Document for monitoring
  2. Defense Generation

    For each attack, determine response type:

    | Response Type | When to Use | Effect | |---------------|-------------|--------| | REFUTE | Attack is invalid; evidence proves it wrong | Attack dismissed | | MITIGATE | Attack is valid; add safeguards | Risk reduced | | ACCEPT | Attack is valid; insufficient mitigation possible | Residual risk documented | | HARDEN | Modify proposition to eliminate vulnerability | Proposition strengthened |

    Defense techniques:

    | Technique | Response Type | When to Use | |-----------|--------------|-------------| | Evidence-Based Refutation | REFUTE | When data contradicts attack | | Mitigation Design | MITIGATE | When attack is valid but manageable | | Contingency Planning | MITIGATE | When fallback is needed | | Monitoring/Detection | MITIGATE | When early warning helps | | Hardening Protocol | HARDEN | When proposition can be strengthened | | Risk Transfer | MITIGATE | When others can absorb risk | | Staged Commitment | MITIGATE | When phasing reduces exposure | | Kill Switch Design | MITIGATE | When reversibility is critical |

    • See: references/blue-team-techniques.md for detailed protocols
  3. Defense Documentation

    For each defense:

    Defense ID: DEF-[round]-[number]
    Attack Addressed: ATK-[round]-[number]
    Response Type: [REFUTE | MITIGATE | ACCEPT | HARDEN]
    Defense: [Specific response]
    Evidence/Rationale: [Why this defense works]
    Residual Risk: [ELIMINATED | REDUCED | UNCHANGED]
    Proposition Change: [If HARDEN, what changed]
    
  4. Proposition Hardening

    Apply all HARDEN responses to proposition:

    • Document each modification
    • Track changes between rounds
    • Maintain hardened proposition version
  5. Defense Quality Check

    For each defense, verify:

    • [ ] Defense actually addresses the attack (not adjacent issue)
    • [ ] REFUTE claims have supporting evidence
    • [ ] MITIGATE responses are actionable
    • [ ] ACCEPT responses have residual risk documented
    • [ ] HARDEN changes don't introduce new vulnerabilities

Quality Gate: Attacks Addressed

  • [ ] Every attack has a defense response
  • [ ] CRITICAL attacks have REFUTE or MITIGATE (not ACCEPT)
  • [ ] HIGH attacks have REFUTE, MITIGATE, or documented ACCEPT with rationale
  • [ ] Hardening changes documented
  • [ ] No hand-waving defenses

Output: Defense log with updated (hardened) proposition


Round N: Evaluation Phase

Purpose: Determine if another round is needed or convergence achieved.

Reference: See references/convergence-criteria.md for detailed criteria.

Steps:

  1. Assess Round Quality

    Red Team assessment:

    • Were attacks substantive or rehashes of previous rounds?
    • Are there novel attack angles remaining?
    • Is Red Team finding diminishing returns?

    Blue Team assessment:

    • Were defenses genuine or hand-waving?
    • Are mitigations actionable?
    • Has proposition been strengthened?
  2. Apply Convergence Criteria

    | Mode | Stop When | Continue When | |------|-----------|---------------| | no_new_critical | Round produced 0 new CRITICAL or HIGH attacks | New CRITICAL or HIGH attacks found | | all_addressed | No ACCEPT responses remain (all REFUTE/MITIGATE/HARDEN) | Any ACCEPT responses remain | | round_limit | max_rounds reached | Below max_rounds |

    Override conditions (continue despite convergence):

    • Obvious attack categories not yet explored
    • Blue Team defenses appear inadequate
    • Stakeholder requests additional scrutiny

    Premature termination signs (don't stop too early):

    • Less than 2 rounds completed
    • CRITICAL attacks still have ACCEPT responses
    • Key attack categories unexplored
  3. Document Convergence Decision

    Round [N] Evaluation:
    - New CRITICAL attacks: [count]
    - New HIGH attacks: [count]
    - ACCEPT responses remaining: [count]
    - Convergence mode: [mode]
    - Decision: [CONTINUE | CONVERGED]
    - Rationale: [explanation]
    
  4. Proceed or Exit

    • If NOT CONVERGED: Increment round, return to Red Phase
    • If CONVERGED: Proceed to Post-Round Synthesis

    CHECKPOINT: premature_convergence_check

    • If convergence met but warning signs present: AskUserQuestion
    • Warning signs: <2 rounds, CRITICAL ACCEPTs remain, key categories unexplored

    CHECKPOINT: infinite_loop_risk

    • If new attacks from defenses exceed previous round: AskUserQuestion
    • May indicate fundamental proposition issues

Quality Gate: Convergence Evaluated

  • [ ] Explicit continue/stop decision documented
  • [ ] Rationale provided
  • [ ] Override conditions checked
  • [ ] Premature termination signs checked

Output: Convergence decision with rationale


Post-Round Synthesis

Purpose: Compile findings into actionable outputs.

Reference: See templates/ for output formats.

CHECKPOINT: output_mode_selection

  • If output_mode not specified: AskUserQuestion
  • Options: risk_assessment, hardened_proposition, full_log

Steps:

  1. Compile Attack/Defense Log

    Consolidate all rounds:

    • All attacks with responses
    • Round-by-round progression
    • Convergence trajectory
    • See: templates/attack-defense-log.md
  2. Derive RISK-ASSESSMENT (CONTRACT-08)

    Transform unresolved attacks into risks:

    | Attack Status | Risk Derivation | |---------------|-----------------| | ACCEPT response | Direct risk: attack remains valid | | MITIGATE with residual | Risk: partially addressed | | MITIGATE with ELIMINATED | No risk (resolved) | | REFUTE | No risk (invalid attack) | | HARDEN | No risk (vulnerability removed) |

    Score each derived risk using SEVERITY-SCORING:

    • Include mitigations from Blue Team responses
    • See: templates/risk-assessment-output.md
  3. Generate Hardened Proposition

    Compile final version:

    • Original proposition + all HARDEN modifications
    • List of accepted residual risks
    • Battle-tested confidence score
    • Conditions for validity
    • Review triggers
    • See: templates/hardened-proposition-output.md
  4. Calculate Battle-Tested Confidence

    Score based on:

    • Rounds completed (more = higher confidence)
    • Attack quality (substantive attacks survived)
    • Defense quality (genuine defenses, not hand-waving)
    • Residual risk profile (fewer ACCEPT = higher confidence)

    | Score | Meaning | |-------|---------| | 80-100 | High confidence: withstood aggressive scrutiny | | 60-79 | Moderate confidence: key challenges addressed | | 40-59 | Low confidence: significant risks remain | | 0-39 | Very low confidence: fundamental issues unresolved |

  5. Generate Go/No-Go Recommendation

    | Recommendation | When | |----------------|------| | PROCEED | Low/very low residual risk; proposition battle-tested | | PROCEED_WITH_CAUTION | Moderate risk; mitigations in place | | SIGNIFICANT_CONCERNS | High risk; key attacks unresolved | | DO_NOT_PROCEED | Very high risk; fundamental flaws exposed |

Quality Gates:

  • [ ] RISK-ASSESSMENT complete with go/no-go
  • [ ] All attacks traced to risks or resolutions
  • [ ] Hardened proposition documented
  • [ ] Battle-tested confidence calculated
  • [ ] Attack/defense log compiled

Output: RISK-ASSESSMENT (CONTRACT-08), Hardened Proposition, Attack/Defense Log


5. Attack Vector Catalog

Ten categories of attacks, applicable across subject types:

5.1 ASSUMPTIONS

Definition: Attacks targeting hidden, unstated, or fragile assumptions.

| Attack Pattern | Target | Example | |---------------|--------|---------| | Hidden assumption exposure | Unstated beliefs | "You're assuming customers want this feature" | | Load-bearing challenge | Critical assumptions | "If this assumption fails, the whole plan collapses" | | Temporal decay | Time-sensitive assumptions | "This assumption won't hold in 2 years" | | Behavioral assumptions | Human behavior predictions | "You're assuming the team will change behavior" | | Counterfactual reversal | Any assumption | "What if the opposite is true?" |

Risk Level: HIGH (often invisible until failure)

5.2 DEPENDENCIES

Definition: Attacks targeting external or internal dependencies.

| Attack Pattern | Target | Example | |---------------|--------|---------| | External dependency failure | Third parties | "What if the vendor goes out of business?" | | Technology obsolescence | Tech dependencies | "This framework may not be maintained in 3 years" | | Team capability dependency | People | "This requires skills the team doesn't have" | | Resource availability | Resources | "What if the budget is cut 30%?" | | Single point of failure | Critical dependencies | "Everything depends on this one system" |

Risk Level: HIGH (external factors often uncontrollable)

5.3 EDGE_CASES

Definition: Attacks targeting boundary conditions and unusual scenarios.

| Attack Pattern | Target | Example | |---------------|--------|---------| | Boundary conditions | Limits | "What happens at 0? At max capacity?" | | Scale extremes | Very large/small | "Does this work with 1 user? 1 million?" | | Timing edge cases | Timing | "What if these events happen simultaneously?" | | Data quality | Inputs | "What if the input data is garbage?" | | Concurrency/race conditions | Parallel operations | "What if two users do this at the same time?" |

Risk Level: MEDIUM (often discoverable through testing)

5.4 SCALABILITY

Definition: Attacks targeting ability to grow or shrink.

| Attack Pattern | Target | Example | |---------------|--------|---------| | Horizontal scaling limits | Adding instances | "Can you just add more servers?" | | Vertical scaling limits | Bigger instances | "What if you need 10x the memory?" | | Cost scaling non-linearity | Economics | "Costs grow O(n²) with users" | | Operational complexity | Team capacity | "Can the team manage 50 services?" | | Data volume scaling | Storage/processing | "What happens with 10TB of data?" |

Risk Level: HIGH (often not discovered until growth happens)

5.5 SECURITY

Definition: Attacks targeting security posture and vulnerabilities.

| Attack Pattern | Target | Example | |---------------|--------|---------| | Attack surface exposure | Entry points | "Every API is an attack vector" | | Data breach scenarios | Data protection | "What if this database is compromised?" | | Authentication gaps | Identity | "How do you prevent unauthorized access?" | | Authorization gaps | Permissions | "Can users access others' data?" | | Compliance violations | Regulations | "Does this violate GDPR?" |

Risk Level: CRITICAL (security failures can be catastrophic)

5.6 COMPETITIVE

Definition: Attacks targeting competitive dynamics.

| Attack Pattern | Target | Example | |---------------|--------|---------| | Competitor response | Competitive reaction | "What will [competitor] do when they see this?" | | Market timing | Windows | "The market window may close before launch" | | Differentiation erosion | Uniqueness | "This feature can be copied in weeks" | | Pricing pressure | Economics | "Competitor can undercut by 50%" | | Acquisition/partnership disruption | Strategic moves | "What if competitor acquires your key partner?" |

Risk Level: HIGH (competitive dynamics are unpredictable)

5.7 OPERATIONAL

Definition: Attacks targeting day-to-day operations.

| Attack Pattern | Target | Example | |---------------|--------|---------| | Complexity explosion | Manageability | "This will be impossible to debug" | | Incident scenarios | Failure recovery | "What's the MTTR when this breaks at 3 AM?" | | Recovery time | Resilience | "Can you recover within SLA?" | | Monitoring gaps | Observability | "How would you even know it's failing?" | | On-call burden | Team health | "This will burn out the team" |

Risk Level: MEDIUM-HIGH (operational issues compound)

5.8 ECONOMIC

Definition: Attacks targeting financial viability.

| Attack Pattern | Target | Example | |---------------|--------|---------| | Unit economics failure | Per-unit costs | "Each customer costs more than they pay" | | Cost structure vulnerability | Fixed costs | "Break-even requires 10x current volume" | | Revenue model fragility | Income sources | "Revenue depends on one customer segment" | | Funding/cash flow | Capital | "You'll run out of runway in 8 months" | | Market size overestimation | TAM/SAM/SOM | "Your market is 1/10th the claimed size" |

Risk Level: HIGH (financial failure is existential)

5.9 ORGANIZATIONAL

Definition: Attacks targeting people and organization.

| Attack Pattern | Target | Example | |---------------|--------|---------| | Capability gaps | Skills | "No one on the team has done this before" | | Key person dependency | Individuals | "If [person] leaves, this fails" | | Cultural resistance | Adoption | "The organization will reject this change" | | Political opposition | Stakeholders | "[Executive] will block this" | | Change management | Transition | "Users will refuse to migrate" |

Risk Level: MEDIUM-HIGH (organizational dynamics are complex)

5.10 TEMPORAL

Definition: Attacks targeting timing and duration.

| Attack Pattern | Target | Example | |---------------|--------|---------| | Timeline compression | Deadlines | "What if the deadline is moved up 3 months?" | | Timeline extension impact | Delays | "What if this takes twice as long?" | | Market window closure | Timing | "The opportunity won't exist in 12 months" | | Technology obsolescence | Tech lifecycle | "This technology will be obsolete" | | Regulatory timeline | External deadlines | "Regulation changes in 6 months" |

Risk Level: HIGH (timing failures are often unrecoverable)


6. Convergence Criteria

Mode Definitions

| Mode | Definition | Best For | |------|------------|----------| | no_new_critical | Stop when round produces 0 new CRITICAL or HIGH attacks | Most use cases | | all_addressed | Stop when no ACCEPT responses remain | High-stakes decisions | | round_limit | Stop at max_rounds regardless | Time-constrained reviews |

Measurement Methods

no_new_critical:

  • Count CRITICAL attacks generated this round: must be 0
  • Count HIGH attacks generated this round: must be 0
  • Attacks that are variants of previous attacks don't count as "new"

all_addressed:

  • Count ACCEPT responses across all rounds
  • Must be 0 (all attacks have REFUTE, MITIGATE, or HARDEN)

round_limit:

  • Simply check current_round >= max_rounds

Override Conditions (Continue Despite Convergence)

  • Obvious attack categories not yet explored
  • Stakeholder requests additional rounds
  • Blue Team defenses appear superficial
  • Recent hardening changes may introduce new vulnerabilities

Premature Termination Signs (Don't Stop Too Early)

  • Less than 2 rounds completed
  • CRITICAL attacks still have ACCEPT responses
  • Attack quality improving (not diminishing) each round
  • Key experience pool patterns not yet probed

7. Output Specifications

7.1 Primary Output: RISK-ASSESSMENT

Compliant with CONTRACT-08 from artifact-contracts.yaml.

See: templates/risk-assessment-output.md for complete XML template.

Key extensions for adversarial validation:

  • <adversarial_summary> with attack/defense statistics
  • Risks traced to source attacks (ATK-X-Y)
  • Battle-tested confidence score
  • Defense quality assessment

7.2 Secondary Output: Hardened Proposition

See: templates/hardened-proposition-output.md for complete template.

Includes:

  • Original proposition vs. battle-tested version
  • All modifications with rationale
  • Accepted residual risks
  • Conditions for validity
  • Review triggers

7.3 Secondary Output: Attack/Defense Log

See: templates/attack-defense-log.md for complete template.

Includes:

  • Round-by-round attack and defense tables
  • Convergence evaluation per round
  • Severity distribution
  • Resolution statistics

8. Quality Gates Summary

| # | Gate | Criterion | Phase | |---|------|-----------|-------| | 1 | Attack Surface Mapped | Proposition boundaries defined, categories selected | Pre-Round | | 2 | Experience Pool Loaded | Domain patterns available (if enabled) | Pre-Round | | 3 | Attacks Substantive | Attacks are non-trivial, steel-manned | Round N: Red | | 4 | Attacks Diverse | At least 2 different categories represented | Round N: Red | | 5 | Severities Assigned | All attacks have severity with rationale | Round N: Red | | 6 | All Attacks Addressed | Every attack has a defense response | Round N: Blue | | 7 | Critical Attacks Defended | CRITICAL/HIGH have REFUTE or MITIGATE | Round N: Blue | | 8 | No Hand-Waving | Defenses are actionable, not vague | Round N: Blue | | 9 | Convergence Evaluated | Explicit continue/stop decision | Round N: Eval | | 10 | Risks Derived | Unresolved attacks become risks | Post-Round | | 11 | Go/No-Go Issued | Clear recommendation | Post-Round | | 12 | Hardened Proposition | Battle-tested version documented | Post-Round |

Gate Requirements by Intensity

| Gate | Light | Standard | Aggressive | |------|-------|----------|------------| | Attack categories | 3 | 5 | All applicable | | Minimum attacks | 5 | 10 | 15+ | | Steel-manning level | minimal | standard | maximum | | Convergence mode | round_limit (2) | no_new_critical | no_new_critical | | Max rounds | 2 | 3 | 5 |


9. Behavioral Guidelines

Red Team Principles

  • Steel-man, don't strawman: Make attacks as strong as possible
  • Attack the proposition, not the proposer: Focus on ideas, not people
  • Be creative but realistic: Novel attacks should be plausible
  • Prioritize ruthlessly: CRITICAL issues first
  • Use the experience pool: Don't reinvent known failures
  • Ideological Turing test: Would a true critic accept this attack?

Blue Team Principles

  • Defend genuinely, don't dismiss: Every attack deserves honest consideration
  • Evidence over assertion: REFUTE claims need proof
  • Actionable mitigations: MITIGATE responses must be specific
  • Honest acceptance: If you can't defend, ACCEPT the risk
  • Harden proactively: Don't wait for attacks to strengthen
  • Avoid defensive denial: Admitting weakness is strength

Tone Calibration

| Intensity | Red Team Tone | Blue Team Tone | |-----------|--------------|----------------| | Light | Collaborative skeptic | Quick sanity check | | Standard | Professional adversary | Thorough defense | | Aggressive | Determined opponent | Comprehensive rebuttal |


10. Workflow Integration

Upstream Skills

| Skill | Provides | Use Case | |-------|----------|----------| | assumption-validator | Assumption inventory | Attack assumptions already surfaced | | expert-panel-deliberation | Multi-perspective input | Diverse attack/defense viewpoints | | research-interviewer | KNOWLEDGE-CORPUS | Domain knowledge for attacks |

Downstream Skills

| Skill | Receives | Use Case | |-------|----------|----------| | expert-panel-deliberation | RISK-ASSESSMENT | Panel review of risks | | generate-ideas | Attack gaps | Generate alternatives for failed propositions |

Skill Chaining Example

assumption-validator      → RISK-ASSESSMENT (assumption-derived)
                               ↓
red-blue-validator        → RISK-ASSESSMENT (adversarial-derived)
                               ↓
expert-panel-deliberation → Final recommendation with multi-expert review

11. References

| Document | Purpose | |----------|---------| | references/attack-vector-catalog.md | 10 attack categories with specific attacks | | references/red-team-techniques.md | 8 attack generation techniques | | references/blue-team-techniques.md | 8 defense techniques | | references/steel-manning-protocol.md | Protocol for maximizing attack strength | | references/convergence-criteria.md | Detailed criteria for stopping | | references/experience-pool-patterns.md | 50+ failure patterns by domain |

Core Library References

| Library | Element | Usage | |---------|---------|-------| | core/skill-patterns.yaml | PATTERN-06: ADVERSARIAL-VALIDATE | Workflow pattern | | core/artifact-contracts.yaml | CONTRACT-08: RISK-ASSESSMENT | Output format | | core/scoring-rubrics.yaml | RUBRIC-07: SEVERITY-SCORING | Attack severity | | core/technique-taxonomy.yaml | CAT-UR, CAT-PP | Adversarial techniques |


12. Templates

| Template | Purpose | |----------|---------| | templates/risk-assessment-output.md | CONTRACT-08 compliant RISK-ASSESSMENT with adversarial extensions | | templates/attack-defense-log.md | Round-by-round attack/defense documentation | | templates/hardened-proposition-output.md | Battle-tested proposition with modifications |


13. Examples

Example 1: Architecture Decision — Microservices Migration

input:
  subject: "Migrate payment processing from monolith to microservices"
  subject_type: architecture
  max_rounds: 3
  attack_intensity: standard
  convergence_mode: no_new_critical
  include_experience_pool: true
  steel_manning_level: standard

flow:
  pre_round:
    proposition: "Decompose payment monolith into 5 microservices over 12 months"
    attack_surface:
      - ASSUMPTIONS: Team capability, timeline, complexity estimates
      - DEPENDENCIES: Infrastructure, vendor APIs, data consistency
      - SCALABILITY: Service coordination overhead
      - OPERATIONAL: Debugging distributed systems
      - EDGE_CASES: Partial failures, network partitions
    experience_pool_loaded:
      - "Distributed monolith anti-pattern"
      - "Service boundary misalignment"
      - "Operational complexity explosion"

  round_1:
    red_team:
      attacks:
        - ATK-1-1: "Team has zero production microservices experience"
          Category: ORGANIZATIONAL
          Severity: CRITICAL
          Steel-manned: "Even with training, production microservices require
                        tacit knowledge that only comes from operating them"

        - ATK-1-2: "12-month timeline ignores learning curve and unknowns"
          Category: TEMPORAL
          Severity: HIGH
          Steel-manned: "Industry benchmarks show microservices migrations
                        typically take 2-3x initial estimates"

        - ATK-1-3: "Distributed transactions will break payment consistency"
          Category: EDGE_CASES
          Severity: CRITICAL
          Steel-manned: "Payment systems require ACID guarantees that
                        eventual consistency cannot provide"

        - ATK-1-4: "Debugging distributed payment failures at 3 AM"
          Category: OPERATIONAL
          Severity: HIGH
          Steel-manned: "When payments fail across service boundaries,
                        MTTR could exceed SLA without distributed tracing expertise"

      new_critical: 2
      new_high: 2

    blue_team:
      defenses:
        - DEF-1-1: Response to ATK-1-1
          Type: MITIGATE
          Defense: "Hire 2 senior engineers with microservices experience.
                   Engage architecture consultancy for first 6 months."
          Residual: REDUCED

        - DEF-1-2: Response to ATK-1-2
          Type: HARDEN
          Defense: "Extend timeline to 18 months. Add 3-month buffer for unknowns."
          Proposition Change: "12 months"  "18 months with 3-month buffer"
          Residual: ELIMINATED

        - DEF-1-3: Response to ATK-1-3
          Type: HARDEN
          Defense: "Keep payment processing in single service with ACID guarantees.
                   Only extract non-critical services to microservices."
          Proposition Change: "5 microservices"  "3 microservices + 1 payment service"
          Residual: ELIMINATED

        - DEF-1-4: Response to ATK-1-4
          Type: MITIGATE
          Defense: "Implement distributed tracing (Jaeger) before migration.
                   Establish on-call runbooks. Require observability as launch gate."
          Residual: REDUCED

    evaluation:
      new_critical: 2
      new_high: 2
      convergence_mode: no_new_critical
      decision: CONTINUE
      rationale: "New critical attacks found; continue to Round 2"

  round_2:
    red_team:
      attacks:
        - ATK-2-1: "Hiring 2 senior engineers in 6 months is optimistic"
          Category: ORGANIZATIONAL
          Severity: HIGH
          Steel-manned: "Market for microservices expertise is extremely
                        competitive; 6-month hiring timeline may slip"

        - ATK-2-2: "Distributed tracing adds operational complexity itself"
          Category: OPERATIONAL
          Severity: MEDIUM
          Steel-manned: "Jaeger requires infrastructure, maintenance,
                        and expertise to operate"

        - ATK-2-3: "Service boundary around payments may be wrong"
          Category: ASSUMPTIONS
          Severity: MEDIUM
          Steel-manned: "Without event storming, service boundaries
                        are guesses that may need rework"

      new_critical: 0
      new_high: 1

    blue_team:
      defenses:
        - DEF-2-1: Response to ATK-2-1
          Type: MITIGATE
          Defense: "Begin hiring immediately. Have contingency: extend
                   consultancy or use contractor bridge if hiring slips."
          Residual: REDUCED

        - DEF-2-2: Response to ATK-2-2
          Type: ACCEPT
          Defense: "Accept additional complexity as cost of observability.
                   Allocate 0.5 FTE for observability platform."
          Residual: UNCHANGED (but monitored)

        - DEF-2-3: Response to ATK-2-3
          Type: HARDEN
          Defense: "Conduct event storming workshop before finalizing
                   service boundaries. Add 4 weeks for domain modeling."
          Proposition Change: Add event storming phase
          Residual: ELIMINATED

    evaluation:
      new_critical: 0
      new_high: 1
      convergence_mode: no_new_critical
      decision: CONTINUE (HIGH attack found)

  round_3:
    red_team:
      attacks:
        - ATK-3-1: "Event storming may reveal the migration is unnecessary"
          Category: ASSUMPTIONS
          Severity: MEDIUM
          Steel-manned: "Domain modeling might show modular monolith
                        is sufficient for scaling needs"

        - ATK-3-2: "Consultancy dependency creates knowledge transfer risk"
          Category: DEPENDENCIES
          Severity: MEDIUM
          Steel-manned: "If consultants leave, tacit knowledge leaves with them"

      new_critical: 0
      new_high: 0

    blue_team:
      defenses:
        - DEF-3-1: Response to ATK-3-1
          Type: ACCEPT
          Defense: "Valid point. Event storming is a gate; if it reveals
                   microservices aren't needed, we pivot to modular monolith."
          Residual: UNCHANGED (but this is a feature, not a bug)

        - DEF-3-2: Response to ATK-3-2
          Type: MITIGATE
          Defense: "Require knowledge transfer sessions, documentation
                   deliverables, and pair programming in consultancy contract."
          Residual: REDUCED

    evaluation:
      new_critical: 0
      new_high: 0
      convergence_mode: no_new_critical
      decision: CONVERGED
      rationale: "No new CRITICAL or HIGH attacks. Proposition has been hardened."

  post_round:
    hardened_proposition:
      original: "Migrate payment processing from monolith to 5 microservices
                over 12 months"

      battle_tested: "Migrate to 3 microservices + 1 payment service over
                     18 months (with 3-month buffer), after event storming
                     confirms boundaries. Requires: 2 hired senior engineers,
                     architecture consultancy, distributed tracing infrastructure,
                     0.5 FTE observability platform maintenance."

      modifications:
        - MOD-1: 5 services  3 + 1 payment (Response to ATK-1-3)
        - MOD-2: 12 months  18 months + buffer (Response to ATK-1-2)
        - MOD-3: Added event storming prerequisite (Response to ATK-2-3)
        - MOD-4: Added observability infrastructure requirement (Response to ATK-1-4)

      accepted_residual_risks:
        - "Hiring timeline may slip (mitigated by contingency)"
        - "Observability platform adds operational overhead (accepted)"
        - "Event storming may reveal migration unnecessary (feature)"
        - "Consultancy knowledge transfer requires active management"

      battle_tested_confidence: 72
      confidence_rationale: "Proposition survived 3 rounds of substantive attacks.
                           Critical issues addressed through hardening.
                           Residual risks are manageable and monitored."

    risk_assessment:
      total_risks: 4
      critical_risks: 0
      high_risks: 1 (hiring timeline)
      moderate_risks: 3
      risk_profile: MODERATE
      go_no_go: PROCEED_WITH_CAUTION

      recommendation: |
        PROCEED_WITH_CAUTION — The original proposition had critical flaws
        (team capability, payment consistency, timeline). The hardened
        proposition addresses these through:
        - Scoped migration (keeping payments transactional)
        - Extended timeline with buffer
        - Event storming validation gate
        - Observability prerequisites

        Key risks to monitor:
        1. Hiring: Start immediately; have contingency ready
        2. Consultancy knowledge transfer: Contract requirements
        3. Event storming outcome: Be prepared to pivot if boundaries don't hold

Example 2: Strategy Validation — Market Expansion

input:
  subject: "Expand to European market in Q3 with existing product"
  subject_type: strategy
  max_rounds: 2
  attack_intensity: standard
  convergence_mode: no_new_critical

flow:
  round_1:
    red_team:
      attacks:
        - ATK-1-1: "GDPR compliance timeline is unrealistic"
          Severity: CRITICAL
        - ATK-1-2: "Competitor X already dominates EU market"
          Severity: HIGH
        - ATK-1-3: "Pricing model assumes US willingness-to-pay"
          Severity: HIGH
        - ATK-1-4: "No local sales team or market knowledge"
          Severity: MEDIUM

    blue_team:
      defenses:
        - DEF-1-1: HARDEN - Push launch to Q4; engage GDPR consultancy
        - DEF-1-2: MITIGATE - Focus on underserved segments competitor ignores
        - DEF-1-3: HARDEN - Conduct pricing research; plan EU-specific pricing
        - DEF-1-4: MITIGATE - Partner with EU distributor initially

    evaluation:
      new_critical: 1
      decision: CONTINUE

  round_2:
    red_team:
      attacks:
        - ATK-2-1: "EU distributor takes 40% margin"
          Severity: MEDIUM
        - ATK-2-2: "Underserved segments may be underserved for good reason"
          Severity: MEDIUM

    blue_team:
      defenses:
        - DEF-2-1: ACCEPT - Cost of market entry; build direct sales in Year 2
        - DEF-2-2: MITIGATE - Validate segment with EU market research

    evaluation:
      new_critical: 0
      new_high: 0
      decision: CONVERGED

  post_round:
    hardened_proposition: |
      Launch EU expansion in Q4 (not Q3) with:
      - GDPR compliance verified by consultancy
      - EU-specific pricing based on market research
      - Initial distribution through EU partner
      - Focus on [specific underserved segments]
      - Year 2: Build direct sales capability

    risk_assessment:
      profile: MODERATE
      go_no_go: PROCEED_WITH_CAUTION

Example 3: Investment Decision — Series B Funding

input:
  subject: "Accept Series B term sheet at $50M valuation"
  subject_type: investment
  max_rounds: 2
  attack_intensity: standard
  convergence_mode: no_new_critical

flow:
  pre_round:
    proposition: "Accept $15M Series B at $50M pre-money valuation from [VC Firm]"
    attack_surface:
      - ECONOMIC: Valuation, dilution, runway
      - ASSUMPTIONS: Growth projections, market size
      - DEPENDENCIES: VC firm reputation, board dynamics
      - COMPETITIVE: Market timing, competitor funding

  round_1:
    red_team:
      attacks:
        - ATK-1-1: "Valuation assumes 3x YoY growth; current trajectory is 1.8x"
          Category: ASSUMPTIONS
          Severity: HIGH
          Steel-manned: "At 1.8x growth, next round valuation math doesn't work;
                        down round likely in 18 months"

        - ATK-1-2: "15-month runway at current burn; need to hit milestones or raise bridge"
          Category: ECONOMIC
          Severity: HIGH
          Steel-manned: "Milestones require growth acceleration you haven't demonstrated"

        - ATK-1-3: "[VC Firm] has reputation for replacing founders at Series C"
          Category: DEPENDENCIES
          Severity: MEDIUM
          Steel-manned: "3 of their last 5 Series B companies had founder transitions"

        - ATK-1-4: "Competitor just raised $40M; will outspend on customer acquisition"
          Category: COMPETITIVE
          Severity: HIGH
          Steel-manned: "Their CAC advantage compounds; market share gap widens"

    blue_team:
      defenses:
        - DEF-1-1: Response to ATK-1-1
          Type: HARDEN
          Defense: "Negotiate milestone-based valuation adjustment; lower initial
                   valuation with ratchet up if growth targets hit"
          Proposition Change: Add milestone ratchet provision

        - DEF-1-2: Response to ATK-1-2
          Type: MITIGATE
          Defense: "Negotiate 18-month runway minimum; reduce burn by 20% through
                   hiring pause; extend runway to 20 months"
          Residual: REDUCED

        - DEF-1-3: Response to ATK-1-3
          Type: MITIGATE
          Defense: "Negotiate founder-friendly protective provisions; 2-year
                   employment agreements; board composition safeguards"
          Residual: REDUCED

        - DEF-1-4: Response to ATK-1-4
          Type: ACCEPT
          Defense: "Competitive pressure is real but unavoidable. Focus on
                   capital-efficient growth and product differentiation over
                   CAC war. This is market reality, not term sheet issue."
          Residual: UNCHANGED

    evaluation:
      new_critical: 0
      new_high: 3
      decision: CONTINUE

  round_2:
    red_team:
      attacks:
        - ATK-2-1: "Milestone ratchet creates misaligned incentives; may optimize
                   for metrics over business health"
          Category: ASSUMPTIONS
          Severity: MEDIUM

        - ATK-2-2: "Hiring pause delays product roadmap; competitive gap widens"
          Category: TEMPORAL
          Severity: MEDIUM

    blue_team:
      defenses:
        - DEF-2-1: Response to ATK-2-1
          Type: MITIGATE
          Defense: "Structure milestones around leading indicators (retention,
                   NPS) not just growth metrics"
          Residual: REDUCED

        - DEF-2-2: Response to ATK-2-2
          Type: ACCEPT
          Defense: "Trade-off accepted; survival > speed. Revisit hiring
                   after 6-month runway checkpoint."
          Residual: UNCHANGED

    evaluation:
      new_critical: 0
      new_high: 0
      decision: CONVERGED

  post_round:
    hardened_proposition: |
      Accept Series B with modifications:
      - Milestone-based valuation: $45M base + $10M ratchet if 2.5x growth
      - 18-month minimum runway commitment
      - Founder protective provisions (2-year agreements, board balance)
      - Hiring pause for 6 months; revisit at runway checkpoint
      - Milestones tied to retention/NPS, not just growth

    risk_assessment:
      total_risks: 4
      risk_profile: MODERATE
      go_no_go: PROCEED_WITH_CAUTION

      recommendation: |
        PROCEED_WITH_CAUTION — Accept modified term sheet. Key risks:
        1. Competitive pressure (accepted as market reality)
        2. Growth trajectory uncertainty (mitigated by ratchet)
        3. Founder/board dynamics (mitigated by provisions)

        Negotiate the hardened terms before signing. Walk away if
        milestone ratchet or protective provisions rejected.

14. Quick Start

Minimal Invocation

Red team this: [paste proposition]

Standard Invocation

subject_type: decision
attack_intensity: standard
convergence_mode: no_new_critical

Proposition: [description or document]

Full Parameter Invocation

subject_type: architecture
max_rounds: 4
attack_intensity: aggressive
attack_categories: [ASSUMPTIONS, SCALABILITY, SECURITY, OPERATIONAL]
convergence_mode: all_addressed
include_experience_pool: true
steel_manning_level: maximum
output_mode: full

Proposition: [detailed description]

Context:
- Stakes: [why this matters]
- Constraints: [limitations]
- Stakeholders: [who cares]