The Agent Skills Directory

[SAFE]: The skill incorporates strong safety guardrails by explicitly instructing the agent not to execute actions without user confirmation ("把决定权留给你", "不擅自执行").- [COMMAND_EXECUTION]: The skill utilizes git status and git diff --stat as diagnostic tools to understand the user's current project state. These are non-destructive, read-only operations that do not involve sensitive data exposure or unauthorized modification.- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources (git output and user dialogue) to generate advice. While this constitutes an attack surface for indirect prompt injection, the impact is mitigated by the structured workflow and the requirement for manual user approval before the agent takes any subsequent steps.

next-step