growth-engine
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Uses the shell command
cutto extract data from a local JSONL file andcurlto send notifications to an external service. It also references a hardcoded absolute local filesystem path/Users/kameronkales/planfi-appin the workflow configuration. - [DATA_EXFILTRATION]: Sends project-specific status updates, including prospect counts and report paths, to a Discord webhook. The webhook URL is retrieved from the agent's memory/configuration rather than being hardcoded.
- [EXTERNAL_DOWNLOADS]: Triggers a workflow that performs automated web searches to identify and qualify potential business prospects.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests data from external web searches and processes it without explicit sanitization or boundary markers.
- Ingestion points:
growth/prospects.jsonland results from thegrowth-engineworkflow (which performs WebSearches). - Boundary markers: Absent in the workflow invocation and file writing steps.
- Capability inventory: File writes (
growth/prospects.jsonl,growth/runs/,growth/monetization-playbook.md), shell execution (cut,curl), and workflow execution. - Sanitization: None identified in the instructions for processing researched data.
Audit Metadata