Skills
Navigate
skills/trailofbits/skills/diagramming-code/Gen Agent Trust Hub

diagramming-code

Pass

Audited by Gen Agent Trust Hub on Oct 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the trailmark command-line tool and a bundled Python script (scripts/diagram.py) via uv run to process source code and output diagram syntax.\n- [EXTERNAL_DOWNLOADS]: The skill installs and uses the trailmark package as its primary analysis engine. This is a recognized tool from the skill author (trailofbits).\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code to generate diagrams, creating an attack surface for indirect prompt injection. This risk is mitigated through documented practices:\n
  • Ingestion points: Source code from the user-specified {targetDir} (as seen in SKILL.md).\n
  • Boundary markers: The agent is instructed in SKILL.md to wrap all generated diagrams in ```mermaid code fences.\n
  • Capability inventory: The skill performs file analysis and shell execution via uv run.\n
  • Sanitization: The references/mermaid-syntax.md file specifies mandatory sanitization of node IDs and label escaping to prevent malformed or malicious output.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 15, 2026, 05:52 PM