alterlab-accreditation-aol
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, user-provided curriculum mapping data through a local Python script (
scripts/aol_matrix.py) to check for structural gaps in accreditation standards. This creates a surface where malicious instructions could be embedded in the processed data. - Ingestion points: The
scripts/aol_matrix.pyscript reads CSV data from a file path provided as a command-line argument or from standard input. - Boundary markers: The instructions do not specify explicit delimiters or warnings for the agent to ignore instructions embedded within the CSV data.
- Capability inventory: The script uses the Python standard library to read files, parse CSV content, and output validation results to the console. It does not perform network operations, execute subprocesses, or write to the file system.
- Sanitization: The script performs structural validation and case normalization (e.g., converting coverage codes to uppercase) but does not include specific sanitization logic to detect or neutralize natural language instructions hidden within the CSV cells.
- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute a local Python script using the
Bashtool (e.g.,uv run python scripts/aol_matrix.py matrix.csv). This is a legitimate function for validating the structure of curriculum matrices but involves direct shell interaction.
Audit Metadata