review-paper
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided document files in various formats (.pdf, .tex, .qmd, .md, .docx), creating a surface where malicious instructions embedded in these documents could influence the agent's behavior.
- Ingestion points: In 'Step 0: Locate and Read the Paper', the skill is instructed to read files directly from paths provided in
$ARGUMENTSor by searching specific local directories including~/Dropbox/Github/[project]/paper/,~/Dropbox/Github/[project]/manuscript/, and~/Dropbox/Github/[project]/draft/. - Boundary markers: The instructions do not define boundary markers or explicit delimiters to separate the document content from the skill's instructions, nor do they include directives to ignore embedded instructions within the papers.
- Capability inventory: The skill is capable of reading from the local filesystem and writing the final report to the working directory. It does not perform network operations, subprocess execution, or arbitrary code evaluation.
- Sanitization: No sanitization, validation, or filtering of the content extracted from the external documents is implemented.
Audit Metadata