research
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to gather data from untrusted third-party sources, which poses a risk of indirect prompt injection.
- Ingestion points: As defined in
SKILL.md, the agent collects 'unfiltered opinions' and 'actual phrases' from platforms such as Reddit, YouTube, and review sites including G2 and Capterra. - Boundary markers: The skill enforces a structured JSON output format which acts as a data boundary, but lacks explicit 'ignore embedded instructions' markers for the extracted content.
- Capability inventory: The
SKILL.mdfile defines a process that outputs JSON to be used by thecampaign-briefskill; no direct code execution or shell commands are invoked within this skill's logic. - Sanitization: The research process described in
SKILL.mddoes not specify requirements for the agent to sanitize or validate the extracted external text for potential prompt injections.
Audit Metadata