paper-writing-section
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is focused on academic writing assistance and contains no patterns of data exfiltration, credential theft, or unauthorized persistence. Its operations are limited to reading local research files and generating LaTeX fragments.
- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection. Ingestion points: It reads user-provided context files ($1), existing LaTeX files, and experiment logs from the workspace. Boundary markers: No specific delimiters are implemented to separate external data from system instructions. Capability inventory: The skill is restricted to text generation; it does not utilize subprocesses, network tools, or file-writing capabilities beyond providing the final LaTeX output. Sanitization: No input filtering is applied to the data ingested from logs or context files. The security risk is classified as negligible due to the skill's restricted capability set.
Audit Metadata