skills/backnotprop/pstack/swarm/Gen Agent Trust Hub

swarm

Pass

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to spawn multiple background subagents, including in the local environment (environment: "local"). Running tasks in the background on the host machine reduces user visibility and could allow for the execution of unauthorized commands or persistent processes if the subagent's brief is manipulated.
  • [INDIRECT_PROMPT_INJECTION]: The skill involves aggregating reports from multiple workers into a final report in Phase C. This creates a vulnerability surface where worker outputs could contain malicious instructions designed to influence the parent agent. The instructions lack explicit boundary markers or sanitization steps.
  • Ingestion points: Phase C (Read the terminal results from workers).
  • Boundary markers: Absent.
  • Capability inventory: Phase B (Spawn workers with subagent_type: generalPurpose, run_in_background: true in cloud/local).
  • Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 19, 2026, 05:44 PM
Security Audit — agent-trust-hub — swarm