follow-up
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The script
followup.pydynamically loads a Python module from a fixed relative path (../pipeline/deals.py) usingimportlib.util. This implementation creates a structural dependency on a neighboring vendor resource from the same author. - [COMMAND_EXECUTION]: The skill's workflow involves executing Python scripts via the command line to check follow-up status, draft messages, and update deal records, which includes both read and write operations on the
closer/pipeline.jsonfile. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data originating from external email replies and contact records (ingestion points:
pipeline.jsonand CLI arguments like--text). This content is interpolated into templates intemplates.jsonwithout the use of explicit boundary markers or sanitization, creating an attack surface where external content might influence the AI agent's message drafting. The skill maintains the capability to write to the filesystem through theP.savefunction.
Audit Metadata