lead-scoring-drift-detector
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing external, potentially untrusted data in the form of JSON or CSV files.
- Ingestion points: The
run.pyscript accepts a file path via the--inputargument and reads its contents into memory usingjson.loads(). - Boundary markers: None identified in the provided scaffold.
- Capability inventory: The script's capabilities are restricted to local file read/write operations within the project directory; no network, command execution, or dynamic evaluation tools are utilized.
- Sanitization: No sanitization or validation of the input data content is performed prior to processing.
- [SAFE]: The provided Python implementation is a standard scaffold that uses built-in libraries (
argparse,json,pathlib) for basic data handling. No malicious patterns such as obfuscation, credential exfiltration, or persistence mechanisms were detected across the skill files.
Audit Metadata