respondent-panel
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user content (the 'artifact') and passes it to multiple synthetic respondent agents for analysis.
- Ingestion points: Untrusted data enters the agent context in
SKILL.mdduring Step 1 ('Get the artifact'), where text, slogans, or landing page content are collected verbatim. - Boundary markers: The instructions emphasize isolation and passing the artifact verbatim, but they lack guidance on using structural delimiters (such as XML tags or triple quotes) to prevent embedded instructions within the artifact from hijacking the sub-agent's persona.
- Capability inventory: The skill orchestrates concurrent calls to the
synthetic-respondentagent, which processes the untrusted artifact. - Sanitization: There is no mention of sanitizing, filtering, or escaping the external content before it is interpolated into the prompts for the synthetic respondents.
Audit Metadata