army-run
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the execution of local shell commands via 'make' to start the supervisor and server components. It also utilizes 'npm' for local package installation and uses 'tmux' for session management.
- [INDIRECT_PROMPT_INJECTION]: The skill functions by orchestrating a fleet of agents that read from a project directory and a database, which creates a surface for indirect prompt injection if the project data contains adversarial content.
- Ingestion points: Local files in the project root defined by 'AOA_PROJECT_ROOT' and entries in the 'aoa.db' SQLite database.
- Boundary markers: The documentation does not specify explicit boundary markers or 'ignore' instructions for data processed by the agents.
- Capability inventory: Subprocess execution via 'acpx', tmux session management, and git worktree manipulation are available to the orchestrated agents.
- Sanitization: The skill documentation focuses on process stability and reconciliation (heartbeats) rather than content-level input sanitization.
Audit Metadata