What makes Self Improvement Llm notable?

1.3K downloads. Trust evidence available

How should Self Improvement Llm be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

Self Improvement Llm

Autonomous AI memory and self-learning system that logs, extracts lessons, verifies improvements, adapts behavior, manages preferences, and generates reusabl... Skill: Self Improvement Llm Owner: brucetangc Summary: Autonomous AI memory and self-learning system that logs, extracts lessons, verifies improvements, adapts behavior, manages preferences, and generates reusabl... Tags: latest:2.3.0 Version history: v2.3.0 | 2026-07-29T11:54:50.774Z | auto Version 2.3.0 - Added a CLI references document (references/cli_ref.md) for command-line usage. - Updated procedural instructio

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.3k

Updated

Oct 10, 2026

Version

2.3.0

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1.3K downloads reported by the source. Last updated 10/10/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 10, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 10, 2026
Adoption signal: 1.3K downloadsadoption · observed Oct 10, 2026
Latest release: 2.3.0release · observed Jul 29, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s17c86x64fn6cy26y1t8rwnn4983f35d:self-improvement-llm

Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.
Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-brucetangc-self-improvement-llm/snapshot"

Documentation

CLAWHUB

145,478 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: self-improvement-llm
description: Autonomous memory and self-learning system for AI agents. Continuously collects experience, manages memory (daily logs, user preferences, knowledge extraction), builds knowledge graphs connecting events→lessons→principles, extracts principles, auto-adjusts behavior, and verifies improvements. Proposes actionable changes for user review before applying. Use when: (1) Agent needs to learn from past sessions, (2) User asks "improve yourself" or "learn from this", (3) Periodic self-evaluation is needed, (4) Agent needs to auto-correct recurring mistakes, (5) Updating AGENTS.md/SOUL.md/MEMORY.md/TOOLS.md based on experience, (6) Extracting universal principles from episodic experiences, (7) Processing user feedback to permanently adapt behavior, (8) Managing daily memory logs, user preferences, or knowledge retention.
---

# Self-Learning System

A continuous learning loop that automatically captures learnings, tracks improvements, and verifies their effectiveness.

**Inspiration:** This skill fuses the structured recording format and detection triggers from [pskoett/self-improving-agent](https://clawhub.ai/pskoett/self-improving-agent) (6.1k installs) with a verification/hypothesis loop that most agent learning systems lack.

## 🤖 Agent: What You Must Do (Not Optional)

These are **required behaviors**, not suggestions. Execute them during conversation without being told.

### During Every Conversation

1. **Detect learning opportunities in real-time.** After every user message or tool result, mentally check:
   - Did the user correct me? → run `python3 scripts/reflect.py --detect "USER_MESSAGE"`
   - Did a tool fail? → log to `.learning-trail.json` immediately
   - Did I learn something new? → log to `.learning-trail.json` immediately

2. **Log corrections immediately.** When the user says "不对", "错了", "no", "actually", etc.:
   ```bash
   python3 scripts/learn.py --log correction "具体纠正了什么"
   ```
   Set pattern_key for dedup: `--log correction "message" --area behavior --priority high`

3. **Log errors automatically.** When a tool call fails or returns unexpected output:
   ```bash
   python3 scripts/learn.py --log error "工具名: 错误简述" --area tooling --priority medium
   ```

4. **After significant tasks,** append to today's daily log:
   ```bash
   python3 scripts/reflect.py --log "完成了什么"
   ```

### At Session Start

5. **Check `.hook-context.txt`** (written by plugin hook at gateway startup):
   Use `read(path="memory/.hook-context.txt")` to check it.
   If it shows pending verifications or patterns ready for promotion, act on them.

6. **Run a quick status check:**
   ```bash
   python3 scripts/learn.py --status
   ```

### During Daily Cycle (via cron, 3AM)

7. The full cycle runs automatically: `python3 scripts/learn.py --cycle`
   - **🌙 Dream:** Distills recent daily logs into MEMORY.md (dedup + compress)
   - Auto-promotes patterns (≥2 occurrences across ≥2 sessions)
   - Auto-generates session summ

_meta.json

{
  "ownerId": "kn7afdk9ag1ftxjn0btmw2tj3h82y30x",
  "slug": "self-improvement-llm",
  "version": "2.3.0",
  "publishedAt": 1785326090774
}

references/cli_ref.md

# CLI Reference & Detailed Structures

## Structured Log Format

Every entry uses this format (inspired by pskoett standard):

### Learning Entry

```
## [LRN-YYYYMMDD-XXX] category:brief_title

**Logged**: ISO-8601 timestamp
**Priority**: low | medium | high | critical
**Status**: pending | in_progress | resolved | wont_fix | promoted
**Area**: frontend | backend | infra | tests | docs | config | behavior | tooling

### Summary
One-line description

### Details
What happened, what was wrong, what's correct

### Suggested Action
Specific fix or improvement

### Metadata
- Source: conversation | error | user_feedback | self_discovery
- Related Files: path/to/file
- Tags: tag1, tag2
- Pattern-Key: unique_key_for_dedup (optional, for recurring patterns)
- Recurrence-Count: 1
- First-Seen: YYYY-MM-DD
- Last-Seen: YYYY-MM-DD
```

### Error Entry

```
## [ERR-YYYYMMDD-XXX] tool_or_command_name

**Logged**: ISO-8601 timestamp
**Priority**: high
**Status**: pending
**Area**: infra | tooling | config

### Summary
Brief description of what failed

### Error
Actual error message or output

### Context
- Command/operation attempted
- Input or parameters used

### Suggested Fix
What might resolve this

### Metadata
- Reproducible: yes | no | unknown
- Related Files: path/to/file
- See Also: ERR-YYYYMMDD-XXX (if recurring)
```

### Feature Request Entry

```
## [FEAT-YYYYMMDD-XXX] capability_name

**Logged**: ISO-8601 timestamp
**Priority**: medium
**Status**: pending
**Area**: as appropriate

### Summary
What the user wanted to do

### User Context
Why they needed it

### Complexity Estimate
simple | medium | complex

### Metadata
- Frequency: first_time | recurring
- Related Features: existing_feature_name
```

### ID Generation

Format: `TYPE-YYYYMMDD-XXX`
- TYPE: LRN (learning), ERR (error), FEAT (feature)
- YYYYMMDD: Current date
- XXX: Sequential number or random 3 chars (e.g., 001, A7B)

## Auto-Generated Skill Format

```markdown
---
name: skill-slug-name
description: 一句话描述这个技能做什么
created: 2026-05-27
updated: 2026-05-27
source: auto
triggers: ["触发关键词或场景"]
tools: [web_fetch, exec, read]
---

## Procedure

1. 步骤一：做了什么
2. 步骤二：怎么做的
3. 步骤三：验证结果

## Pitfalls

- 已知问题或陷阱
- 容易出错的地方
- 环境依赖

## Verification

- 如何验证结果正确
- 预期输出是什么
```

## Verification Loop — JSON Entry

```json
{
  "id": "change-20260505-001",
  "source": "LRN-20260505-003",
  "target": "TOOLS.md",
  "change": "Added 'prefer read over exec for files'",
  "hypothesis": "This will reduce file-viewing errors",
  "verified": false,
  "next_check": "2026-05-12",
  "evidence": []
}
```

### Verification Outcomes

| Result | Action |
|--------|--------|
| ✅ Confirmed effective | Mark verified, reduce monitoring to monthly |
| ❌ Ineffective | Revert change, log why it failed |
| ❌ Made worse | Revert immediately, escalate |
| ❓ Inconclusive | Extend monitoring, add more data points |

## Conflict Resolution — Priority Score

```
Score = BasePriority(100/60/30/10) + RecurrenceBonus(×10 each) + RecencyBonus(

references/reflection_frameworks.md

# Reflection Frameworks

Deep-dive reference for types of reflection, depth levels, score rubrics, proposal patterns, and the learning system.

## Learning System Architecture

The self-learning system runs as a background process, not an on-demand command:

```
                    ┌──────────┐
                    │  AGENT    │
                    │  (LLM)    │
                    └────┬─────┘
                         │
         ┌───────────────┼───────────────┐
         │               │               │
         ▼               ▼               ▼
   ┌──────────┐   ┌──────────┐   ┌──────────┐
   │ Session  │   │  Memory  │   │ Learning │
   │  Log     │   │  Files   │   │  Trail   │
   └──────────┘   └──────────┘   └──────────┘
         │               │               │
         └───────────────┼───────────────┘
                         ▼
                  ┌──────────────┐
                  │  Heartbeat   │
                  │  (Idle)      │
                  └──────┬───────┘
                         │
                         ▼
                  ┌──────────────┐
                  │ Learn Cycle  │
                  │  extract →   │
                  │  verify →    │
                  │  integrate   │
                  └──────────────┘
                         │
                         ▼
                  ┌──────────────┐
                  │  Self-Modify │
                  │  (files)     │
                  └──────────────┘
```

### Data Flow

1. **Session Logging** — After each task, auto-append to `memory/YYYY-MM-DD.md`
2. **Learning Trail** — `memory/.learning-trail.json` tracks every change, its hypothesis, and verification status
3. **Heartbeat** — During idle time, triggers `python3 scripts/learn.py --cycle`
4. **Verify** — Checks if past changes actually improved behavior (measured by error rate)
5. **Adapt** — Reverts failed changes, reinforces successful ones

### Auto-Logging Format

```markdown
### ✅ 14:32 - Fetched weather data for Rugao
### ❌ 14:35 - Tried to send screenshot via exec
   Error: Platform requires MEDIA directive, not curl
```

Three lines max per entry. Keep it scannable.

### Learning Trail Structure

```json
{
  "changes": [
    {
      "id": "change-20260505-001",
      "target": "TOOLS.md",
      "hypothesis": "Adding MEDIA note prevents file delivery failures",
      "verified": false,
      "next_check": "2026-05-12"
    }
  ],
  "watchlist": [
    {"issue": "Using exec instead of read for files", "count": 3, "status": "watch"}
  ]
}
```

## Industry Patterns

These are the real-world patterns used by mature agent frameworks:

### 1. Reflexion (Academic, 388⭐)
**Paper:** Shinn et al., NeurIPS 2023 — [arXiv:2303.11366](https://arxiv.org/abs/2303.11366)
**Official Code:** [noahshinn/reflexion-draft](https://github.com/noahshinn/reflexion-draft)

```
┌──────────┐    task + reflections    ┌──────────────┐
│  Actor   │ ─────────────────────► │     LLM       │
└──────────┘ ◄──────────────────── └──────────────┘
     │

hooks/openclaw/HOOK.md

---
name: self-improvement
description: Self-learning system — checks pending learnings, writes session context, and tracks patterns at gateway startup
metadata:
  openclaw:
    emoji: "🧠"
    events: ["gateway:startup"]
---

# Self-Improvement Gateway Hook

Runs at gateway startup. Writes actionable context to `memory/.hook-context.txt` for the agent to read at session start.

## What It Does

- Checks `.learning-trail.json` for pending high-priority items
- Checks for overdue verifications
- Detects patterns ready for promotion (≥2 occurrences)
- Checks if recent session summaries exist
- Writes findings to `memory/.hook-context.txt`

## Agent Usage

At session start, the agent should:
```bash
cat memory/.hook-context.txt
```

This file is regenerated at each gateway startup and contains the current state of the learning system.

## Installation

The hook is auto-installed by OpenClaw when the skill is enabled.

Github ReposUpdated 1d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.