guideline-generation

Pass

Audited by Gen Agent Trust Hub on Feb 24, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • External Data Ingestion (Indirect Prompt Injection Surface): The skill is designed to process external materials such as PDFs, transcripts, and discovery reports. This establishes a surface where instructions embedded within these documents could potentially influence the agent's behavior during the guideline generation process.\n
  • Ingestion points: Processes documents (PDF, PPTX, DOCX, MD, TXT) and conversation transcripts as specified in SKILL.md.\n
  • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are defined for the parsing phase.\n
  • Capability inventory: The skill has the capability to write and rename files in the .claude/ directory.\n
  • Sanitization: The skill includes explicit instructions in the 'Privacy and Security' section to redact PII and anonymize sensitive information detected during processing.\n- File System Persistence: The skill maintains state by saving and archiving guideline files within the user's working directory (.claude/brand-voice-guidelines.md). While this is a common pattern for agent continuity, it involves local file write and rename operations.\n- Delegation to Sub-Agents: The workflow involves passing data to specialized agents (document-analysis, conversation-analysis). This modular approach is functional but extends the data processing chain across multiple internal components.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 24, 2026, 06:31 AM