message-market-fit
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and analysis of external content from user-provided URLs (such as company websites, competitor pages, and advertisement copy) to perform its messaging audit. This process introduces a surface for indirect prompt injection where the agent could potentially process instructions embedded in the analyzed content.
- Ingestion points: SKILL.md Step 2 (Collect Assets + Competitors) requests URLs for websites and ads.
- Boundary markers: No specific delimiters or safety instructions are defined to wrap the retrieved strings during the audit process.
- Capability inventory: The skill uses the agent's web-fetching capabilities to retrieve textual data for comparative analysis and scoring.
- Sanitization: The instructions do not specify sanitization or filtering of the fetched external text.
- [DATA_EXFILTRATION]: The skill performs network operations to retrieve content from external domains specified by the user during the audit phase. This is a functional requirement for auditing existing marketing assets and benchmarking against competitors.
- [SAFE]: All instructions are directed toward the legitimate primary purpose of the skill (product marketing strategy). The steering instructions (e.g., "Confirm before advancing", "Push back on weak inputs") are designed to ensure high-quality output and pacing rather than to override safety protocols.
Audit Metadata