Learning System
ReviewAudited by ClawScan on May 10, 2026.
Overview
The core learning-note workflow is coherent, but the skill can scan local memory/work notes and send weekly summaries through Feishu with unclear recipient and approval boundaries.
Review this skill before installing if your OpenClaw memory or notes contain sensitive work. The local health and mastery scripts look purpose-aligned, but do not enable `--quick`, weekly cron, or Feishu sending until you have confirmed the exact recipient, message content, and approval behavior.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Personal learning history, research notes, and work context in the OpenClaw workspace may become persistent context for future tasks.
The workflow reads local agent memory logs and stores learning notes that may later be indexed and reused as context.
读取本周的 memory 日志（最近 7 天） ... 学习笔记存在 PARA 的 areas/ 下，自动被 memory_search 索引
Use this only for notes you are comfortable making persistent; keep secrets, private customer data, and sensitive work details out of the scanned memory/notes paths.
If used casually, the agent may update persistent learning records or knowledge-map ratings without another review checkpoint.
Quick mode intentionally removes confirmation steps while the workflow can create notes and update the knowledge map.
`--quick`: 跳过确认节点，全自动执行
Avoid `--quick` until you trust the workflow; review generated note and knowledge-map changes before accepting them.
A weekly learning summary could expose private work, research, or PR context to a Feishu destination the user did not intend.
The weekly-review guide combines local memory/PR extraction with sending a summary through Feishu, and it names a specific recipient rather than clearly scoping the destination to the installing user.
从每日日志中提取：...代码改动和 PR... 通过飞书发送给青雲，包含：- 本周学习亮点 ... - 下周重点方向
Before enabling weekly review or any Feishu integration, confirm the recipient, message contents, credentials used, and whether the agent must ask before sending.
If configured, the skill can run recurring reviews that read local notes/memory and may generate or send summaries without a fresh manual command each week.
The README recommends an optional scheduled job that can trigger the learning review periodically.
配置每周回顾 Cron ... 每周日晚自动触发学习回顾
Only configure the cron job if you want recurring automation, and pair it with explicit review/approval before external sending.