using-atm
Pass
Audited by Gen Agent Trust Hub on Jun 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to utilize several command-line tools for orchestration and session management, including
atm,am(Agent Mail),tmux,git, andgh. - The skill specifically uses
tmux send-keysandatm sendto inject commands and prompts into active tmux panes representing worker agents. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes external data from a task queue.
- Ingestion points: Data enters the context via the
bd readycommand (picking beads from a queue). - Boundary markers: None identified in the instructions for delimiting bead content.
- Capability inventory: The skill has access to shell execution (
atm,am,tmux), git operations, and GitHub CLI commands. - Sanitization: No specific sanitization or validation of the bead content is mentioned before it is dispatched to agent panes.
- [EXTERNAL_DOWNLOADS]: References a dependency on
atm, which is described as a local binary fork of the Named Tmux Manager (ntm) located at~/dev/ntm/dist/atm-darwin-arm64.
Audit Metadata