@fxri/toolkit@1.17.0
专为多人 + AI 跨项目协作打造:任务管理与配套的规范沉淀、规则、技能分发为主,附带多语言 CHANGELOG 与文档站。
AI security review
scanned · by lpm-firewall-aiThe earlier block is supported: CLI startup automatically replaces conflicting agent skill entries, and the replacement path recursively deletes the existing entry. This is a reachable agent-control hijack under the package's skill-management purpose, so the appropriate intent class is dangerous_capability.