sdo-project-setup

Pass

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes project-specific data from external files which could serve as a surface for indirect instructions.
  • Ingestion points: The skill reads PLAN.md to parse YAML frontmatter and interacts with existing .docx files to update metadata.
  • Boundary markers: PLAN.md uses standard triple-dash (---) delimiters for metadata blocks.
  • Capability inventory: The skill utilizes file system access to copy templates, write to .docx files, and execute helper scripts via uv run.
  • Sanitization: The metadata syncing script set_docx_props.py uses the lxml library for XML generation, which ensures that property values are properly escaped when written to the document XML.
  • [COMMAND_EXECUTION]: The skill documentation and prompts instruct the agent to use shell commands such as cp for project initialization and uv run to execute the Python-based metadata sync tool.
  • [DATA_EXFILTRATION]: The set_docx_props.py script accesses the USERNAME environment variable. This is used solely to populate the 'Author' and 'Creator' fields in the document's core properties, which is a standard functional requirement for generating document metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 9, 2026, 02:34 AM