Self Improvement

🦾 Self-Improving Agent (Strict Promotion Protocol)

"Stop letting your AI agent pollute its own core instructions based on subjective feelings."

This is an industrial-grade, closed-loop Self-Improvement System for OpenClaw. It replaces the default subjective "agent feelings" with a rigid, quantitative pipeline: The Rule of 3.

Instead of bloating your SOUL.md or AGENTS.md with every random bug the agent encounters, this skill forces the agent to merely log errors as pending. Only when an error recurs 3 times (linked via See Also) is it eligible for promotion.

Includes a built-in bash script (promote-review.sh) to aggregate pending promotions for human approval.

🌟 Why install this?

1. Zero Context Bloat: Keeps your SOUL.md and 200K window pristine.
2. Rule of 3 (Quantitative Trigger): Issues must happen 3 times before they can be considered a "Rule".
3. Strict Domain Governance: Mutually exclusive rules for what goes into SOUL (behavior), AGENTS (workflow), and TOOLS (CLI gotchas).
4. Human-in-the-Loop Review: Batch processing of promotions via a weekly aggregator script.

🚀 Quick Setup (OpenClaw)

This skill is designed natively for the OpenClaw architecture.

Via ClawdHub CLI (1-Click Install):

clawhub install self-improving-agent

Workspace Structure

OpenClaw injects these files into every session:

~/.openclaw/workspace/
├── AGENTS.md          # Multi-agent workflows, delegation patterns
├── SOUL.md            # Behavioral guidelines, personality, principles
├── TOOLS.md           # Tool capabilities, integration gotchas
├── MEMORY.md          # Long-term memory (main session only)
├── memory/            # Memory System (v4.0)
│   ├── daily_raw/
│   ├── summaries/
│   ├── projects/
│   └── core/          # This skill's log files
│       ├── learning.md
│       ├── error.md
│       └── features.md

Create Learning Files

mkdir -p ~/.openclaw/workspace/memory/core

Then create the log files (or copy from assets/):

• learning.md — corrections, knowledge gaps, best practices
• error.md — command failures, exceptions
• features.md — user-requested capabilities

Promotion Targets

When learnings prove broadly applicable, promote them to workspace files:

Inter-Session Communication

OpenClaw provides tools to share learnings across sessions:

• sessions_list — View active/recent sessions
• sessions_history — Read another session's transcript
• sessions_send — Send a learning to another session
• sessions_spawn — Spawn a sub-agent for background work

Optional: Enable Hook

For automatic reminders at session start:

# Copy hook to OpenClaw hooks directory
cp -r hooks/openclaw ~/.openclaw/hooks/self-improvement

# Enable it
openclaw hooks enable self-improvement

See references/openclaw-integration.md for complete details.

Generic Setup (Other Agents)

For Claude Code, Codex, Copilot, or other agents, create memory/core/ in your project:

mkdir -p memory/core

Copy templates from assets/ or create files with headers.

Logging Format

Learning Entry

Append to memory/core/learning.md:

## [LRN-YYYYMMDD-XXX] category

**Logged**: ISO-8601 timestamp
**Priority**: low | medium | high | critical
**Status**: pending
**Area**: frontend | backend | infra | tests | docs | config

### Summary
One-line description of what was learned

### Details
Full context: what happened, what was wrong, what's correct

### Suggested Action
Specific fix or improvement to make

### Metadata
- Source: conversation | error | user_feedback
- Related Files: path/to/file.ext
- Tags: tag1, tag2
- See Also: LRN-20250110-001 (if related to existing entry)

---

Error Entry

Append to memory/core/error.md:

## [ERR-YYYYMMDD-XXX] skill_or_command_name

**Logged**: ISO-8601 timestamp
**Priority**: high
**Status**: pending
**Area**: frontend | backend | infra | tests | docs | config

### Summary
Brief description of what failed

### Error

Actual error message or output

### Context
- Command/operation attempted
- Input or parameters used
- Environment details if relevant

### Suggested Fix
If identifiable, what might resolve this

### Metadata
- Reproducible: yes | no | unknown
- Related Files: path/to/file.ext
- See Also: ERR-20250110-001 (if recurring)

---

Feature Request Entry

Append to memory/core/features.md:

## [FEAT-YYYYMMDD-XXX] capability_name

**Logged**: ISO-8601 timestamp
**Priority**: medium
**Status**: pending
**Area**: frontend | backend | infra | tests | docs | config

### Requested Capability
What the user wanted to do

### User Context
Why they needed it, what problem they're solving

### Complexity Estimate
simple | medium | complex

### Suggested Implementation
How this could be built, what it might extend

### 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)

Examples: LRN-20250115-001, ERR-20250115-A3F, FEAT-20250115-002

Resolving Entries

When an issue is fixed, update the entry:

1. Change **Status**: pending → **Status**: resolved
2. Add resolution block after Metadata:

### Resolution
- **Resolved**: 2025-01-16T09:00:00Z
- **Commit/PR**: abc123 or #42
- **Notes**: Brief description of what was done

Other status values:

• in_progress - Actively being worked on
• wont_fix - Decided not to address (add reason in Resolution notes)
• promoted - Elevated to CLAUDE.md, AGENTS.md, or .github/copilot-instructions.md

严格晋升机制 (Strict Promotion Protocol)

CRITICAL RULE: The main agent is explicitly FORBIDDEN from writing directly to SOUL.md, AGENTS.md, or TOOLS.md during normal tasks. All new knowledge must be recorded in memory/core/*.md first.

When to Trigger Promotion (Rule of 3)

We utilize a strict quantitative counter rather than subjective feelings.

1. Log the Issue: If an issue occurs, log it to memory/core/ (status: pending).
2. Search Historical Logs: ALWAYS search memory/core/ for similar past issues using grep.
3. Trigger Threshold: ONLY if the current issue links to 3 or more existing entries via **See Also**, you MUST change the status of the current entry to **Status**: ready_for_promotion.
4. Wait for Review: Do not write to workspace files. Leave it as ready_for_promotion for the weekly human review pipeline.

Strict Domain Separation (Target Governance)

If the weekly review script approves a promotion, it must strictly follow these mutually exclusive rules:

Promotion Examples

Learning (Accumulated 3+ times):

Project uses pnpm workspaces. Attempted npm install but failed 3 times across sessions.

In AGENTS.md (After human approved ready_for_promotion):

## Build Protocol
- MANDATORY: Use `pnpm install`, NOT npm, for workspace root syncing.

Recurring Pattern Detection

If logging something similar to an existing entry:

1. Search first: grep -r "keyword" memory/core/
2. Link entries: Add **See Also**: ERR-20250110-001 in Metadata
3. Bump priority if issue keeps recurring
4. Consider systemic fix: Recurring issues often indicate:
   • Missing documentation (→ promote to CLAUDE.md or .github/copilot-instructions.md)
   • Missing automation (→ add to AGENTS.md)
   • Architectural problem (→ create tech debt ticket)

Periodic Review

Review memory/core/ at natural breakpoints:

When to Review

• Before starting a new major task
• After completing a feature
• When working in an area with past learnings
• Weekly during active development

Quick Status Check

# Count pending items
grep -h "Status\*\*: pending" memory/core/*.md | wc -l

# List pending high-priority items
grep -B5 "Priority\*\*: high" memory/core/*.md | grep "^## \["

# Find learnings for a specific area
grep -l "Area\*\*: backend" memory/core/*.md

Review Actions

• Resolve fixed items
• Promote applicable learnings
• Link related entries
• Escalate recurring issues

Detection Triggers

Automatically log when you notice:

Corrections (→ learning with correction category):

• "No, that's not right..."
• "Actually, it should be..."
• "You're wrong about..."
• "That's outdated..."

Feature Requests (→ feature request):

• "Can you also..."
• "I wish you could..."
• "Is there a way to..."
• "Why can't you..."

Knowledge Gaps (→ learning with knowledge_gap category):

• User provides information you didn't know
• Documentation you referenced is outdated
• API behavior differs from your understanding

Errors (→ error entry):

• Command returns non-zero exit code
• Exception or stack trace
• Unexpected output or behavior
• Timeout or connection failure

Priority Guidelines