beads-orchestrate
Warn
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The orchestrator dynamically constructs and executes prompts for implementation agents via the Agent tool. Prompts are constructed at runtime by interpolating data from external sources, such as bead descriptions and hand-off notes, into a predefined template.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external bead notes and agent reports which can influence the orchestrator's planning and command execution.\n
- Ingestion points: Untrusted content enters the context via
bd showcommand results and relayed reports from sub-agents.\n - Boundary markers: Instructions explicitly state that agent reports are claims and not user approval, requiring the orchestrator to treat them as unverified output.\n
- Capability inventory: The orchestrator has the ability to execute shell commands, manage git worktrees, and launch additional sub-agents.\n
- Sanitization: The skill requires the orchestrator to manually verify agent claims against the repository state (e.g.,
git log, CI status) before relaying information or closing beads.\n- [COMMAND_EXECUTION]: The skill uses the Bash tool to perform repository management tasks, including git worktree creation, dependency installation (e.g.,npm ci), and environment cleanup (e.g.,ddev delete).
Audit Metadata