dayz-proxy-align

Warn

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/install_py3d.py uses subprocess.run to execute shell commands. Specifically, it calls pip install and pip uninstall to manage the py3d package environment. While it attempts to validate the package using AST analysis, the ability to mutate the system's package state is a significant capability.
  • [DYNAMIC_EXECUTION]: Multiple scripts employ dynamic code loading techniques:
  • scripts/proxy_viewer.py uses importlib.util.spec_from_file_location and module_from_spec to locate and execute viewer_core.py by traversing parent directories. This could allow for the execution of arbitrary code if a file named viewer_core.py is planted in a parent directory.
  • scripts/worn_overlay.py modifies sys.path using sys.path.insert(0, debin_scripts) where debin_scripts is a path provided via a command-line argument. It then imports modules (odol_reader, odol_to_mlod) from this external directory.
  • scripts/install_py3d.py uses importlib.import_module('py3d') to dynamically inspect the package before modification.
  • [INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection:
  • Ingestion points: The skill processes .p3d binary files and recipe.json structured data in scripts/proxy_extract.py, scripts/proxy_canon.py, scripts/proxy_viewer.py, and scripts/worn_overlay.py.
  • Boundary markers: There are no explicit boundary markers or warnings to the agent regarding embedded instructions in these external data formats.
  • Capability inventory: The skill possesses capabilities for filesystem writes (scripts/proxy_apply.py), shell command execution (scripts/install_py3d.py), and dynamic code loading.
  • Sanitization: While it uses standard JSON parsing, the data extracted (such as point indices and coordinates) is used directly to index internal structures and modify files without secondary validation against malicious payloads.
  • [ENVIRONMENT_PROBING]: The scripts/install_py3d.py script contains hardcoded fallback glob patterns targeting specific environment paths (/sessions/*/mnt/*/_tools/py3d/dist/*.whl). This indicates the skill is tailored for a specific, non-standard execution environment and probes for resources in shared system locations.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 8, 2026, 08:46 AM