菜单
Generic agent self-improvement skill built on OpenClaw-RL research (arxiv.org/abs/2603.10165). Captures evaluative signals (+1/-1) and directive hints from any user feedback, generates improvement suggestions.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | self-improvement |
| description | Generic agent self-improvement skill built on OpenClaw-RL research (arxiv.org/abs/2603.10165). Captures evaluative signals (+1/-1) and directive hints from any user feedback, generates improvement suggestions. |
| metadata | {"openclaw":{"emoji":"🧬","homepage":"https://your-domain.example.com"}} |
Enables any agent to learn and improve from user feedback using PRM-style evaluation.
From OpenClaw-RL paper:
"Next-state signals encode both evaluative and directive information about the preceding action."
Two signal types:
SKILL:self-improvement --feedback "Great!" --job daily-report
SKILL:self-improvement --feedback "Add more stats" --job daily-report
SKILL:self-improvement --stats daily-report
SKILL:self-improvement --improve daily-report
| Arg | Description | Example |
|---|---|---|
--job | Task/job name | daily-digest, weekly-recap |
--feedback | User response | "Thanks!", "Add more X" |
--score | Manual score override | 1, 0, -1 |
--stats | Show feedback stats | daily-digest |
--improve | Generate improvements | daily-digest |
--weekly | Weekly summary | (flag) |
memory/learning/agent-feedback.json:
{
"jobs": {
"daily-digest": {
"evaluations": [
{ "date": "2026-03-14", "score": 1, "hint": null },
{ "date": "2026-03-13", "score": -1, "hint": "add weekly star comparison" }
],
"improvements": [
{ "date": "2026-03-14", "suggestion": "Add weekly star delta", "implemented": false }
]
}
}
}
{
"id": "feedback-capture",
"schedule": "30 9 * * *",
"message": "SKILL:self-improvement --job daily-digest"
}
{
"id": "improvement-suggestions",
"schedule": "0 10 * * *",
"message": "SKILL:self-improvement --improve daily-digest → Telegram"
}
{
"id": "weekly-summary",
"schedule": "0 9 * * 6",
"message": "SKILL:self-improvement --improve all --weekly"
}
📈 Improvement Suggestions - {job}
Stats: 8 evaluations, avg score: 0.75
Top Hints:
1. "add weekly star changes" (2x)
2. "use table format" (1x)
Suggested Actions:
• Add 7-day star delta to GitHub section
• Use table-image-generator for stats
Status: 1 improvement pending
If no user feedback available, manually evaluate:
Evaluate yesterday's output as: good/bad
Orchestrate multi-agent teams with defined roles, task lifecycles, handoff protocols, and review workflows. Use when: (1) Setting up a team of 2+ agents with different specializations, (2) Defining task routing and lifecycle (inbox → spec → build → review → done), (3) Creating handoff protocols between agents, (4) Establishing review and quality gates, (5) Managing async communication and artifact sharing between agents.
Look up any arxiv paper on alphaxiv.org to get a structured AI-generated overview
Socratic questioning protocol + user communication. MANDATORY for complex requests, new features, or unclear requirements. Includes progress reporting and error handling.
Lighthouse-style efficiency audit for OpenClaw. Scores your instance A+ to F across 6 categories (context injection, cron health, session bloat, config, skills, transcripts). Identifies wasted tokens, bloated sessions, misconfigured crons, and model right-sizing opportunities. Zero dependencies (Python stdlib only).
Make your AI agent learn and improve automatically. Reviews sessions, extracts learnings, updates memory files, and compounds knowledge over time. Set up nightly review loops that make your agent smarter every day.
A multi-agent deliberation hub with 3 core agents and extensible extended agents. Can call user workspace skills when needed.