dwg-extraction

Fail

Audited by Gen Agent Trust Hub on Mar 17, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The compile_libredwg.sh script performs a git clone from https://github.com/LibreDWG/libredwg.git during execution. This external repository is not part of the trusted vendors list.
  • [REMOTE_CODE_EXECUTION]: The skill downloads C source code and compiles it at runtime using gcc into an executable binary (dwg2dxf). This binary is then executed to process user-provided DWG files. This pattern of downloading, building, and running code from an external source at runtime is a high-risk activity.
  • [COMMAND_EXECUTION]: The skill uses bash to run compilation scripts, gcc to produce the converter tool, and python3 to execute the entity parser. It also invokes the compiled binary via the command line for file conversion.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. 1. Ingestion point: parse_dxf.py processes data from user-supplied DWG files converted to DXF. 2. Boundary markers: No delimiters or "ignore embedded instructions" warnings are used when processing the extracted text. 3. Capability inventory: The skill has the capability to execute subprocesses, run gcc, and perform file system operations. 4. Sanitization: The parser only performs regex-based cleaning of CAD formatting and does not validate the semantic content of the extracted text before it enters the agent context.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Mar 17, 2026, 09:54 PM
Security Audit — agent-trust-hub — dwg-extraction