attack-vector-generation git:20260519.dc15fe2A

Immutable. This exact content is served forever at /api/v1/blob/f375171c596b37c9.

---
name: attack-vector-generation
description: Generate specific attack strategies for a given threat surface, producing concrete probes that can be executed.
execution: subagent
prompt: ./prompt.md
input: surface (string), artifact (string), prior_findings (string)
used-by: [red-teaming]
---

# Attack Vector Generation

Generates concrete, executable attack vectors targeting a specific threat surface.

## Execution

Subagent — spawned via subagent-spawning/spawn-agent.

## Why Subagent

Vector generation requires creative adversarial thinking without defensive contamination. Each surface needs fresh attack ideation.

## Input

- **surface**: The specific threat surface to target (from threat-surface-mapping)
- **artifact**: The artifact being attacked
- **prior_findings**: Results from previous probes (to avoid repetition)

## Output

- **vectors**: List of specific attack vectors with description, expected outcome, and severity estimate
- **priority_order**: Recommended execution order (highest-impact first)
- **follow_up_triggers**: Conditions that should trigger deeper investigation