tao-validate-recipe-transfer

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The script scripts/preprocess_parity.py uses importlib.import_module to dynamically load and execute functions from local Python modules specified via the --path command-line argument. This is intended for comparing user-defined preprocessing functions but allows for the execution of arbitrary code within the local environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including COCO-format JSON annotations and image files, through its reporting scripts (domain_gap_report.py, split_leakage_check.py). The resulting reports are intended to be read by the AI agent, which could lead to indirect prompt injection if the source data contains malicious instructions designed to influence the agent's behavior.
  • Ingestion points: scripts/domain_gap_report.py (COCO JSON), scripts/split_leakage_check.py (Image metadata/filenames), scripts/preprocess_parity.py (YAML/JSON specs).
  • Boundary markers: Not explicitly present in the generated report templates.
  • Capability inventory: The skill uses Read, Bash, and Write tools to execute Python scripts, read files, and generate markdown/HTML reports.
  • Sanitization: The scripts use standard JSON/YAML parsing; preprocess_parity.py uses yaml.safe_load() for spec files.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 02:56 AM
Security Audit — agent-trust-hub — tao-validate-recipe-transfer