Skill Extractor

从复杂任务中提取可复用的技能文档，参考 Hermes Agent 的 Skill Documents 设计。用于当任务完成后，识别值得保留的步骤流程，并存储为可搜索的技能文件，下次遇到类似任务时自动检索并建议使用。

This skill appears coherent and localized: it analyzes conversation text, generates SKILL.md files, and stores/indexes them under ~/.openclaw/workspace/skills/skill-extractor/. Before using/installing, consider: (1) privacy — conversations (including any secrets) may be written to disk in plaintext; if you handle sensitive data, do not enable automatic saving or consider encrypting that directory; (2) review/confirm saves — prefer manual confirmation rather than automatic "save" so you control what gets persisted; (3) runtime requirements — the search script uses SQLite FTS5 (your Python's sqlite must have FTS5 enabled) and there are minor implementation issues (e.g., ORDER BY rank) that may cause runtime errors; (4) audit usage — because the agent can invoke the script autonomously, limit or monitor automated suggestions if you don't want the agent to store conversational data without explicit approval. Overall the code is local and does not attempt network exfiltration or request unrelated credentials.

Type: OpenClaw Skill Name: niuniu-skill-extractor Version: 1.0.0 The bundle implements a 'Skill Extractor' utility designed to help an AI agent learn from its own conversation history by saving successful workflows as markdown files. The scripts (extract_skill.py, search_skills.py, suggest_skill.py) use standard Python libraries for regex-based parsing, local file management in the user's home directory, and SQLite FTS5 for indexing. No evidence of data exfiltration, remote code execution, or malicious prompt injection was found; the code logic is transparent and strictly aligned with the stated purpose of workflow documentation.