pengyouquan-pangyu

Use when user mentions "胖鱼朋友圈" and wants to write a WeChat Moments post. Analyzes user's past writing style from reference materials, guides through a conversational process to understand sharing needs, then generates style-consistent copy for user confirmation. Triggered by keywords like 胖鱼朋友圈, 发胖鱼朋友圈, 写胖鱼朋友圈.

This skill legitimately stores example posts and a short style summary, but its runtime instructions will: (1) automatically append incoming 'reference' text into references/posts.md (often without asking), and (2) perform file writes before sending a confirmation to the user. Before installing, consider whether you are comfortable with the agent automatically persisting anything you paste or send as a 'reference'. Ask where references/ will be stored (local vs. cloud) and who can access them. If you need stricter privacy, request that the skill be changed to always ask for explicit consent before saving, and to confirm with the user before performing any persistent write. Also ensure the platform's 'file write' tool is trusted and that stored files are not exposed to unintended parties.

Type: OpenClaw Skill Name: pengyouquan-pangyu Version: 1.0.2 The skill is designed to help users write WeChat Moments posts by learning their writing style. It explicitly instructs the AI agent to read from and write to `references/posts.md` and `references/style.md` within its own skill bundle directory. These file operations are for accumulating user posts and generating style summaries, which is a legitimate function for a stateful, learning skill. There is no evidence of data exfiltration, malicious execution, persistence mechanisms, or prompt injection attempts aimed at subverting the agent for harmful purposes. The 'CRITICAL' instructions in SKILL.md are focused on ensuring the agent reliably performs the intended local file operations, not on malicious activities.