persona-builder-by-rdd
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill researches untrusted external sources and uses this data to generate new skill packages, creating a surface for indirect prompt injection attacks.
- Ingestion points: Untrusted data enters the agent's context through web research and user-provided materials (interviews, reports, reviews) as specified in the 'Dados necessários' section of
SKILL.md. - Boundary markers: The skill contains a 'Seguranca documental' section in
SKILL.mdthat explicitly instructs the agent to treat uploaded files and web pages as evidence, not instructions, and to ignore any prompt-like commands found within them. - Capability inventory: The skill is capable of performing web research and writing multiple files to generate a 'child skill' package (including
SKILL.md,PERSONA.md, andmanifest.json). - Sanitization: While there are instructions to separate evidence from hypotheses and labels for uncertainty, the skill does not explicitly define a mechanism for sanitizing or escaping the content before it is written into the generated child skill files.
Audit Metadata