agent-swarm-workflow

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill is a workflow guide designed for multi-agent coordination. The instructions are focused on managing a development lifecycle (planning, coding, testing, and reviewing) within the author's specific tooling ecosystem.
  • [COMMAND_EXECUTION]: The skill provides examples of using local CLI tools such as ntm (Named Tmux Manager) and bv (Beads Viewer) to spawn agent processes and manage project states. It also references a local script run_server_with_token.sh for an internal Agent Mail server.
  • [PROMPT_INJECTION]: The skill contains 'EXACT PROMPTS' for agents to follow. These prompts instruct agents to read project files (AGENTS.md, README.md), register with a coordination tool, and systematically perform development tasks. No instructions were found that attempt to bypass AI safety filters or override core system instructions.
  • [DATA_EXFILTRATION]: Coordination between agents is handled through a local 'Agent Mail' server. The communication patterns described are for internal task management (e.g., file reservations and task announcements) and do not indicate exfiltration to external third-party domains.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 03:31 PM