run-plan
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to load and execute instructions from implementation plans, which introduces a surface for indirect prompt injection if the plan content is malicious.
- Ingestion points: Loads external plan files provided in the workspace (Step 1).
- Boundary markers: Enforces a strict information barrier for the validator subagent using a fixed prompt template to prevent the main agent from leaking worker context or implementation strategies.
- Capability inventory: Spawns worker and validator subagents, modifies files, executes shell commands, and handles git commits.
- Sanitization: Relies on critical manual/agent review steps and binary-verdict validation cycles rather than automated string sanitization.
- [COMMAND_EXECUTION]: The skill executes shell commands for project discovery, task verification, and end-to-end testing as specified in the plan.
- Evidence: Step 0 and Step 3 involve identifying and running verification commands and test suites to validate project state and task implementation.
Audit Metadata