skills/a5c-ai/babysitter/anti-drift/Gen Agent Trust Hub

anti-drift

Pass

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates hierarchical coordination where agents share state and process external data via WebFetch and WebSearch. This creates a surface for indirect prompt injection where malicious instructions embedded in web content or shared memory could influence agent behavior.
  • Ingestion points: Shared state from swarm agents and outputs from WebFetch and WebSearch tools specified in SKILL.md.
  • Boundary markers: The instructions do not define delimiters or specific safety markers to differentiate between coordination data and instructions.
  • Capability inventory: The skill allows access to powerful tools including Bash, Write, Edit, and Agent, providing a significant capability surface if injection occurs.
  • Sanitization: No input validation or sanitization mechanisms are defined in the coordination workflow.
  • [COMMAND_EXECUTION]: The SKILL.md file instructs the agent to execute a coordination script located at 'methodologies/ruflo/ruflo-swarm-coordination' to manage swarm checkpoints and alignment validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 24, 2026, 04:45 AM
Security Audit — agent-trust-hub — anti-drift