executing-plans
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
scripts/task-doneutility executes arbitrary shell commands provided as arguments (the test command). This is a primary functionality designed to verify code changes against expected outputs during the implementation loop. - [INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting and executing instructions from external implementation plans and task briefs. While this creates an injection surface, the skill includes explicit guardrails.
- Ingestion points: Implementation plan files and dynamically generated task briefs in the project workspace.
- Boundary markers: Instructions specify that the agent must stop and consult the human user for any security-sensitive actions, destructive operations, or side effects outside the worktree.
- Capability inventory: Shell command execution (
scripts/task-done), project workspace management, and Git operations. - Sanitization: Relies on agent adherence to the "Four things stop you" safety protocol rather than automated sanitization of plan content.
Audit Metadata