refactor-structure

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a bundled script, scripts/outline_extract.py, to parse LaTeX documents. Analysis confirms the script is safe as it relies solely on the Python standard library, performs no network operations, and is limited to read-only file access with safeguards against recursive file inclusion depth.\n- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by ingesting and processing content from user-provided LaTeX files. \n
  • Ingestion points: User-provided LaTeX source files and sub-files (via \input or \include) processed in Step 1 of SKILL.md.\n
  • Boundary markers: No explicit delimiters are used when the extracted structure is presented to the agent, though the script limits extraction to headings and first sentences.\n
  • Capability inventory: Execution of local Python scripts and the ability to modify local files (move/merge/split/cut) upon user approval.\n
  • Sanitization: The outline_extract.py script performs basic LaTeX macro stripping for readability but does not implement security-specific sanitization of the ingested text.\n- [SAFE]: The skill includes strong guardrails requiring user verification of all findings and approval of all document changes. It does not perform any remote code execution, exfiltrate data, or bypass safety guidelines.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 06:16 PM
Security Audit — agent-trust-hub — refactor-structure