coordinator
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill constructs prompts for sub-agents by reading untrusted data from project files and conversation history.
- Ingestion points: The skill reads project markdown files (such as plans and specifications) and previous conversation context to write prompt files for sub-agents.
- Boundary markers: It uses quoted shell heredocs (EOF) when writing temporary files to delimit the instructions, which helps prevent unintended variable expansion.
- Capability inventory: The skill utilizes the Bash tool for orchestration and the Write tool to generate prompt files.
- Sanitization: External content is interpolated verbatim into prompts, relying on the safety guardrails of the sub-agents.
- [COMMAND_EXECUTION]: The skill relies on the Bash tool to interface with the workmux CLI for its core functionality. This includes managing agent lifecycles and the ability to execute arbitrary shell commands within the worktree panes of managed agents using the workmux run command.
Audit Metadata