competitive-brief
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8) due to its core function of processing untrusted data from external sources.\n
- Ingestion points: The skill instructs the agent to research competitor websites, blogs, social media profiles, and third-party review sites such as G2, Reddit, and community forums (SKILL.md).\n
- Boundary markers: There are no explicit instructions or delimiters in the prompt to separate external research data from the agent's internal logic, nor are there warnings to ignore instructions found within external content.\n
- Capability inventory: The agent uses Read, Glob, and Grep tools and generates comprehensive analysis reports, battlecards, and messaging recommendations based on the findings (SKILL.md).\n
- Sanitization: The skill lacks validation or sanitization of ingested web content before it is incorporated into the structured analysis output.
Audit Metadata