skills/a5c-ai/babysitter/bug-bounty/Gen Agent Trust Hub

bug-bounty

Pass

Audited by Gen Agent Trust Hub on Nov 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and validate external security reports and Proof-of-Concept (PoC) code, which creates a surface for indirect prompt injection. Malicious researcher-provided input could attempt to influence the agent's behavior during the triage process.
  • Ingestion points: Untrusted security reports and PoC code submissions as referenced in the triage and validation sections of SKILL.md and README.md.
  • Boundary markers: The skill does not provide explicit delimiters or instructions to ignore or isolate embedded directions within researcher-provided reports.
  • Capability inventory: The skill utilizes the Bash tool to execute command-line operations such as 'git clone' and 'forge test', which provides a path for potential code execution if PoCs are malicious.
  • Sanitization: No specific mechanisms for sanitizing or pre-scanning researcher-provided PoC code before execution in a test environment are documented.
  • [EXTERNAL_DOWNLOADS]: The skill references established security research tools and documentation from reputable organizations.
  • Evidence: Links to Trail of Bits and mark3labs GitHub repositories are provided in SKILL.md for security analysis and static analysis tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 5, 2026, 02:11 PM