extracting-code-structure

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the extraction of metadata (such as docstrings, comments, and structure) from source code files, which creates a surface for indirect prompt injection if the processed files are untrusted.
  • Ingestion points: The agent ingests data from local source files via the ts-pack process <file> command as documented in SKILL.md.
  • Boundary markers: The skill does not specify boundary markers or instructions to the agent to treat the extracted metadata (especially comments and docstrings) as untrusted content.
  • Capability inventory: The skill demonstrates capabilities for executing shell commands (ts-pack, jq, git) and using a Python SDK (tree_sitter_language_pack) to handle file data.
  • Sanitization: No sanitization or validation logic is provided to filter out potential instructions embedded within the source files before the agent processes the resulting JSON metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 09:04 AM
Security Audit — agent-trust-hub — extracting-code-structure