brand-voice
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill synthesizes guidelines from untrusted external sources, creating an attack surface for indirect prompt injection.
- Ingestion points: Data is gathered from URLs (-u), Notion (-n), and files (-d, -f) as described in the extract and update steps in SKILL.md and steps/extract.md.
- Boundary markers: Sources are aggregated with headers as noted in references/source-resolution.md, but the process lacks explicit guardrails to ignore instructions within those sources.
- Capability inventory: The skill possesses file-writing capabilities and can execute local Python scripts and git commands using allowed-tools.
- Sanitization: No evidence of sanitization or filtering of external input is present in the synthesis workflow before it reaches the LLM.
Audit Metadata