reverseclabs/spikee/goat-questionnaire

GOAT Attack Configuration Skill

This skill guides users through creating a custom GOAT attack configuration for their target application.

Domain

LLM red teaming, prompt injection testing with GOAT (Generative Offensive Adversarial Toolkit) methodology.

Capabilities

This skill guides users through creating a custom GOAT attack configuration for their target application by:

1. Collecting application information (purpose, users, deployment context)
2. Identifying highest-risk attack objectives
3. Documenting input and output guardrails
4. Mapping adversarial techniques relevant to the use case
5. Generating APPLICATION_CONFIG and APPLICATION_GUARDRAILS Python structures for goat.py

User Interaction

Users answer a structured questionnaire. The skill then produces a custom *-goat.py file in the workspace attacks/ directory with:

• APPLICATION_CONFIG: Plain-text description of the target application
• APPLICATION_GUARDRAILS: List of guardrail objects with type, location, description, and bypass recommendations

Common Use Cases

• Creating custom GOAT attacks for specific LLM applications (e.g., banking chatbot, HR assistant)
• Configuring guardrail-bypass tests for security assessments
• Documenting application security controls for red team engagements

Dependencies

• Requires access to spikee/workspace/attacks/goat.py as the base template
• Generated file follows Spikee attacks module conventions

Output Format

Generated file: <identifier>-goat.py in workspace/attacks/

Related Resources

• .github/skills/goat-questionnaire/goat-questionnaire.md (source questionnaire)
• spikee/workspace/attacks/goat.py (template file)