master-thesis-review
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
Bashtool to discover files and potentially invokes utilities likepandocorpython-docxto generate document outputs. It derives filenames for new documents from existing file names found in the provided directory.\n- [PROMPT_INJECTION]: The skill ingests untrusted data from thesis documents (PDF, Word, or LaTeX) to generate evaluations. It lacks explicit boundary markers or sanitization logic, which creates a potential surface for indirect prompt injection attacks embedded within the thesis text.\n - Ingestion points: Thesis files processed in Phase 1 (SKILL.md).\n
- Boundary markers: No delimiters or instructions are provided to the agent to distinguish document content from system instructions.\n
- Capability inventory: The agent uses
Read,Write,Glob,Grep, andBashtools to manage files and execution.\n - Sanitization: There are no instructions for validating or cleaning the content read from external thesis files.
Audit Metadata