framing-doc
Pass
Audited by Gen Agent Trust Hub on Aug 15, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from external transcript files.
- Ingestion points: The
SKILL.mdfile instructs the agent to "Get specific file paths" and "Read them in the order the user specifies." - Boundary markers: The instructions lack explicit syntax-level delimiters (like XML tags or specific block markers) for the transcript content, although they require logical separation in the final output (Source vs. Interpretation).
- Capability inventory: The skill uses file reading capabilities to process transcripts. No network operations, file writing, or command execution tools are utilized or requested in the frontmatter.
- Sanitization: There are no explicit instructions to sanitize, escape, or filter the content of the transcripts before they are processed by the LLM.
Audit Metadata