credo-read-then-forget
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze external, untrusted data in the form of papers, preprints, and PDFs. This creates a surface for indirect prompt injection where malicious instructions embedded in a document could attempt to hijack the agent's behavior.
- Ingestion points: Defined in the description and body for processing "a paper, preprint, or PDF" provided by the user or found in the environment.
- Boundary markers: The instructions lack delimiters or explicit warnings to the agent to ignore any natural language instructions found within the analyzed documents.
- Capability inventory: The skill utilizes the agent's ability to read and parse document files, though it does not explicitly invoke additional tools like network or file system writes within this specific skill file.
- Sanitization: There is no requirement for the agent to sanitize or filter the content of the papers before analysis.
Audit Metadata