offers
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is composed of static markdown documentation and contains no scripts, binary files, or automated system commands. No evidence of obfuscation, credential harvesting, or unauthorized persistence was found.
- [INDIRECT_PROMPT_INJECTION]: The skill includes instructions to ingest information from local project files (e.g., product-marketing.md) to provide relevant advice. While this creates a surface for indirect prompt injection, the skill does not possess any dangerous capabilities to exploit this data. 1. Ingestion points: Reads context from .agents/product-marketing.md, .claude/product-marketing.md, or product-marketing-context.md. 2. Capability inventory: The skill defines no tools and contains no scripts for shell execution, file modification, or network access. 3. Boundary markers: The skill provides instructions to the agent to limit its questioning based on existing context. 4. Sanitization: Content is processed via the agent's standard conversational interface.
Audit Metadata