prospecting
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted external data from sources such as public forums (Reddit, Hacker News), GitHub issues, and prospect websites to identify demand signals and qualifying criteria.
- Ingestion points: Browser-based research, community forum mining, and automated website content extraction (via Firecrawl/Browserbase).
- Boundary markers: The instructions explicitly mandate paraphrasing external content and using minimal quotations to mitigate the risk of the agent following instructions embedded in scraped data.
- Capability inventory: The agent is authorized to execute shell commands, perform network lookups via integrated B2B tools, and generate CSV or HTML output files.
- Sanitization: The framework includes instructions to escape all values when generating standalone HTML reports to prevent cross-site scripting or other injection issues.
- [COMMAND_EXECUTION]: The skill utilizes a local Node.js script (
tools/clis/github-prospects.js) to automate the enrichment of lead data from GitHub repositories. - [SAFE]: The skill implements a robust compliance section addressing GDPR, CAN-SPAM, CASL, and the Terms of Service for major platforms (LinkedIn, Google Maps), ensuring that prospecting activities remain within legal and ethical boundaries.
Audit Metadata