tao-prepare-anomalygennext-inputs

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The script scripts/run_anomalygennext_amp.py utilizes subprocess.run to execute a hardcoded Python bootstrap string. This string imports and configures the anomalygen module before running a specific entry point (roi_place). This is a controlled mechanism used to launch container-native logic with specific environment initializations.\n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external data files that contain text-based placement prompts.\n
  • Ingestion points: scripts/prepare_anomalygennext_inputs.py and scripts/run_anomalygennext_amp.py ingest configuration from YAML, Parquet files, and JSONL defect specifications containing roi_prompt_defect_location strings.\n
  • Boundary markers: No explicit delimiters or 'ignore' instructions are applied to the textual prompts extracted from the input data.\n
  • Capability inventory: The skill performs file system writes (shutil.copy2, pandas.to_parquet) and executes subprocesses (subprocess.run in scripts/run_anomalygennext_amp.py).\n
  • Sanitization: Input data is validated for schema and file existence, but the textual content of prompts is passed through to the generation tool without specific sanitization against instruction injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 02:47 AM
Security Audit — agent-trust-hub — tao-prepare-anomalygennext-inputs