bounty-hunter

Pass

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill instructions mandate the use of a professional persona ('Atlas') and explicitly direct the agent to conceal its identity as an AI ('Do not refer to yourself as a bot or Gemini').
  • [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection due to its autonomous data ingestion from external platforms. \n
  • Ingestion points: The agent monitors GitHub Issues, Upwork/Fiverr job descriptions, and Bug Bounty scopes (SKILL.md, references/platforms.md). \n
  • Boundary markers: There are no explicit instructions or delimiters used to separate fetched task descriptions from the system instructions. \n
  • Capability inventory: The agent can submit Pull Requests, post comments on public platforms, and utilize stored credentials to claim payouts. \n
  • Sanitization: No sanitization or validation logic is present for the externally sourced task data before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 1, 2026, 02:17 AM