office-files
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill converts user-supplied Word documents (
theirs.docx) into Markdown for the agent to read and edit. This creates an attack surface where instructions hidden within the document could be executed by the agent upon reading the converted content. - Ingestion points: Processes
theirs.docxusingpandocto generatedoc.md(SKILL.md). - Boundary markers: Absent; the instructions do not include delimiters or warnings to treat the document content as untrusted text.
- Capability inventory: Shell execution (
bash), Python execution (python3), and file system writes (SKILL.md). - Sanitization: Absent; no sanitization is performed on the Markdown produced from user documents.
- [DYNAMIC_EXECUTION]: The skill uses dynamic execution patterns to perform document manipulation and environment setup.
- Evidence: Uses
python3 -cfor inline execution and `python3 - <<'PY'` heredocs to run multi-line Python scripts for document assembly (SKILL.md).
- [COMMAND_EXECUTION]: The skill relies on shell commands to install dependencies and execute document conversion utilities.
- Evidence: Use of
bashsnippets forpip installandpandocfile conversion (SKILL.md).
Audit Metadata