billing-forecast-analyzer
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input data from a local file, creating a surface for potential instruction injection through the data payload.
- Ingestion points: The run.py script (line 32) reads data via Path(args.input).read_text() and json.loads().
- Boundary markers: No delimiters or ignore instructions are used for the input JSON content.
- Capability inventory: The script performs file writes to the output path (line 35).
- Sanitization: No input validation is performed on the data content.
Audit Metadata