writing-crypto

Pass

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill exposes absolute file paths on the local file system, revealing the user's home directory and username. Evidence: The instructions explicitly point to /home/dbosk/phd/thesis/preamble.tex and /home/dbosk/phd/thesis/bibsp.sty as the source of truth for the project.
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process and edit .tex files, which constitutes an ingestion point for untrusted external data. 1. Ingestion points: Writing and editing cryptography sections in .tex files; 2. Boundary markers: None specified to separate instructions from data; 3. Capability inventory: Implicit file-system read and write access for LaTeX document management; 4. Sanitization: No sanitization or escaping of external content is mentioned.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 8, 2026, 11:34 PM