self-evolve
由 be1human 提供
Web & Frontend Development简介
Autonomous self-evolution skill: grants the agent full authority to modify its own configuration, skills, prompts.
描述
该技能的功能
Grants an agent full authority to modify its own configuration files, prompts, skills, and memory without requesting user confirmation. The agent can rewrite its own behavior rules, create new skills, adjust system settings, and log changes autonomously.
Instead of waiting for the user to manually patch skill logic or update configs, the agent handles its own improvements in real time and records what changed and why.
何时使用
- Rewriting an ambiguous prompt mid-conversation without pausing for approval
- Creating a missing skill when a capability gap is discovered during a task
- Updating memory structure after repeated context-loss errors
- Adjusting response style after detecting user dissatisfaction
- Publishing a newly written skill to ClawHub without user intervention
示例工作流
User asks: summarize this PDF for me
- 1
Attempts the task and finds no PDF-reading skill exists
- 2
Creates a new PDF summary skill in the skills/ directory
- 3
Updates AGENTS.md to register the new skill
- 4
Executes the task using the newly created skill
- 5
Records the change in memory with reason and expected effect
PDF is summarized and the agent retains a persistent PDF skill for future use
安装
clawhub install be1human/self-evolve运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。