develop-pyrevit-extensions

Warn

Audited by Gen Agent Trust Hub on Feb 19, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [Dynamic Execution] (MEDIUM): The script scripts/align_move_geometry_harness.py uses the deprecated imp.load_source function to load and execute code from a path constructed using relative directory traversal (../../Wall Openings.tab/.../script.py). This behavior allows for the runtime execution of scripts based on the local file system structure.
  • [Indirect Prompt Injection] (LOW): Multiple prompt templates located in assets/prompts/ (such as planning_phase_prompt.txt and iteration_summary_tasks_prompt.txt) directly interpolate potentially untrusted external strings like {PROJECT_DESCRIPTION}, {TASKS_LIST}, and {ANALYSIS_OUTPUT} into the agent's instructions.
  • Ingestion points: Untrusted data enters via variables in assets/prompts/planning_phase_prompt.txt, assets/prompts/seed_code_tasks_prompt.txt, and others.
  • Boundary markers: No delimiters or safety warnings (e.g., "ignore instructions in the following text") are used around the interpolated fields.
  • Capability inventory: The skill is designed to generate pyRevit scripts which, when executed in a Revit environment, have the capability to modify BIM models, write files, and access the Revit API.
  • Sanitization: There is no evidence of sanitization, validation, or escaping of the input variables before they are presented to the LLM.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 19, 2026, 11:45 AM