Security audit
Security checks for vulnerabilities and agentic risk
The skill openly automates multi-repository coding work, but it can read projects, write plan files, start a persistent loop, and dispatch worker CLIs without a clear confirmation checkpoint.
Install only if you intentionally want an agent to read selected repositories, create or update planning files, start a background dashboard and loop, and dispatch coding CLI workers. Use a narrow project root, review generated plans before allowing workers to run, and make sure you know how to stop it with the documented down command.
The trigger phrases are broad enough that ordinary user requests about '推进项目', '后台常驻', or '自动推进计划' could unintentionally activate a powerful orchestration skill. In this skill’s context, accidental invocation is more dangerous because activation leads directly to repository scanning, plan generation, file writes, and background loop startup without an explicit confirmation gate.
This section instructs unattended ingestion and startup of a persistent background loop after writing or updating repository files, while explicitly telling the controller not to ask follow-up questions. That combination removes informed consent for actions that modify project files, launch long-running services, and dispatch downstream agents, increasing the risk of unintended changes, resource consumption, and hard-to-stop automation.
No suspicious patterns detected.
