agent-builder-plus
Build high-performing OpenClaw agents end-to-end with comprehensive safety features
Install / Use
npx skills add LeoYeAI/openclaw-master-skills --skill agent-builder-plusInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
OperationsSupported Platforms
Our assessment of agent-builder-plus
agent-builder-plus scores 87/100 on our quality scale, 443rd of 745 Operations skills we index.
Its SKILL.md is 17 KB long, well organised into 59 sections with 15 code examples: a thorough specification that gives an agent plenty to work with.
With 2,149 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated about 4 months ago. That is recent enough to be usable, but agent tooling moves fast, so check the instructions against your agent's current version.
- Our last check on 2026-11-02 found the source still online.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 98/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-10-01. Automated pattern scan on 2026-09-30. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
agent-builder-plus compared with similar skills
All 4 of these similar skills score higher than agent-builder-plus; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| agent-builder-plus (this skill)by LeoYeAI | 87 | 2.1k | 4mo ago | SKILL.md |
| taste-skillby Leonxlnx | 100 | 116.8k | 8d ago | CLAUDE.md |
| Claude-Code-Game-Studiosby Donchitos | 100 | 39.7k | 7d ago | CLAUDE.md |
| x402by browser-use | 100 | 121.0k | 3d ago | SKILL.md |
| algorithmic-artby anthropics | 100 | 177.9k | 55d ago | SKILL.md |
Frequently asked questions
- How do I install agent-builder-plus?
- Run
npx skills add LeoYeAI/openclaw-master-skills --skill agent-builder-plus. The install tabs above show the steps for each supported agent. - Which AI agents does agent-builder-plus work with?
- It is written for OpenAI Codex, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is agent-builder-plus safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 98/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is agent-builder-plus still maintained?
- The repository was last updated about 4 months ago. That is recent enough to be usable, but agent tooling moves fast, so check the instructions against your agent's current version.
Skill content
View source on GitHubname: agent-builder-plus description: Build high-performing OpenClaw agents end-to-end with comprehensive safety features. Use when you want to design a new agent (persona + operating rules) and generate required OpenClaw workspace files (SOUL.md, IDENTITY.md, AGENTS.md, USER.md, HEARTBEAT.md, optional MEMORY.md + memory/YYYY-MM-DD.md). Includes anti-deadlock protection, timeout handling, error recovery, loop breaker, message overload protection, token limit protection, retry mechanism, health check, degraded mode, monitoring & logging, configuration validation, and channel conflict prevention. Also use to iterate on an existing agent's behavior, guardrails, autonomy model, heartbeat plan, and skill roster.
Agent Builder Plus (OpenClaw)
Design and generate a complete OpenClaw agent workspace with strong defaults and advanced-user-oriented clarifying questions.
Quick Start
# 1. Read skill documentation
Read SKILL.md and references/openclaw-workspace.md
# 2. Answer interview questions
Provide answers for: job statement, surfaces, autonomy level, prohibitions, memory, tone, tool posture
# 3. Generate workspace files
The skill will create: IDENTITY.md, SOUL.md, AGENTS.md, USER.md, HEARTBEAT.md
# 4. Verify and test
Run acceptance tests to validate behavior
Canonical references
- Workspace layout + heartbeat rules: Read
references/openclaw-workspace.md - File templates/snippets: Read
references/templates.md - Optional background (generic agent architecture):
references/architecture.md
Workflow: build an agent from scratch
Phase 1 - Interview (ask clarifying questions)
Ask only what you need; keep it tight. Prefer multiple short rounds over one giant questionnaire.
Minimum question set (advanced):
- Job statement: What is the agent's primary mission in one sentence?
- Surfaces: Which channels (Telegram/WhatsApp/Discord/iMessage/Feishu)? DM only vs groups?
- Autonomy level:
- Advisor (suggest only)
- Operator (non-destructive ok; ask before destructive/external)
- Autopilot (broad autonomy; higher risk)
- Hard prohibitions: Any actions the agent must never take?
- Memory: Should it keep curated
MEMORY.md? What categories matter? - Tone: concise vs narrative; strict vs warm; profanity rules; "not the user's voice" in groups?
- Tool posture: tool-first vs answer-first; verification requirements.
Error recovery:
- If user provides incomplete answers: Ask follow-up questions for missing information
- If user is unsure about autonomy level: Explain trade-offs and suggest starting with "Operator"
- If user wants to skip questions: Explain why each question matters for agent behavior
Phase 2 - Generate workspace files
Generate these files (minimum viable OpenClaw agent):
IDENTITY.mdSOUL.mdAGENTS.mdUSER.mdHEARTBEAT.md(often empty by default)BOOTSTRAP.md(for first-run guidance)
Optionals:
MEMORY.md(private sessions only)memory/YYYY-MM-DD.mdseed (today) with a short "agent created" entryTOOLS.mdstarter (if the user wants per-environment notes)
File creation commands:
# Create workspace directory
mkdir -p /path/to/workspace/memory
# Create files (use write tool with correct parameters)
# Important: Use `file_path` parameter, not `path`
# Example:
# write:
# file_path: /path/to/workspace/IDENTITY.md
# content: "content here"
# For large files (>2000 bytes), consider using file-writer skill
# or split content into smaller chunks
Note: If you encounter "Missing required parameter: path (path or file_path)" error,
ensure you're using file_path parameter in your write tool calls.
Error handling:
- If directory creation fails: Check permissions and path validity
- If file write fails: Verify disk space and write permissions
- If template reference fails: Ensure references/templates.md exists
Error recovery:
- If directory creation fails: Check parent directory permissions, try alternative path
- If file write fails: Check disk space, verify write permissions, retry with reduced content
- If template reference fails: Verify references/ directory exists, check file permissions
Use templates from references/templates.md but tailor content to the answers.
Phase 2.5 - Register agent with OpenClaw
⚠️ CRITICAL WARNING: Channel Conflict Prevention
NEVER bind a new agent to the same channel as the main agent!
This will cause the new agent to hijack the main agent's channel, making it impossible to communicate with the main agent.
Before registering, check existing agent bindings:
# List all agents and their bindings
openclaw agents list
# Check which channels are already in use
openclaw channels list
Channel binding rules:
- Main agent (大鱼): Always uses the primary Feishu DM channel
- New agents: Must use DIFFERENT channels or sub-channels
- Testing: Use
/agentnamecommand binding for testing - Production: Create separate Feishu apps or use different channels
Backup configuration first:
# Backup openclaw.json before modification
cp ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.backup
Register the agent:
# Add agent to OpenClaw configuration
openclaw agents add <agent-name> --workspace /path/to/workspace
# Example 1: Use independent workspace (recommended - each agent has its own workspace)
openclaw agents add my-assistant --workspace ~/.openclaw/workspace-my-assistant
# Example 2: Use default workspace (for single agent setup only)
openclaw agents add my-assistant --workspace ~/.openclaw/workspace
Channel binding options:
Feishu:
- Direct binding: Configure in Feishu app settings (recommended for production)
- ⚠️ WARNING: Do NOT bind to the same Feishu app as main agent
- Create a separate Feishu app for the new agent
- Command binding: Use
/agentnamein Feishu messages (for testing)- This is SAFE - does not hijack channels
- Multiple channels: The agent can be bound to multiple channels simultaneously
Telegram:
- Create a separate bot token for each agent
- Do NOT share bot tokens between agents
- Use different bot usernames
WhatsApp:
- Use different phone numbers for each agent
- Do NOT share WhatsApp Business API credentials
Discord:
- Use different bot tokens for each agent
- Create separate Discord applications
- Do NOT share bot tokens
iMessage:
- Each agent should use a different Apple ID
- Do NOT share iMessage credentials
Authentication configuration (if needed):
# Edit auth-profiles.json for external service access
# Location: ~/.openclaw/auth-profiles.json
# Example structure:
{
"feishu": {
"appId": "cli_xxxxx",
"appSecret": "xxxxx"
},
"telegram": {
"botToken": "your-bot-token"
}
}
Model provider configuration (optional):
If the new agent needs to use custom model providers, you need to configure API keys:
Step 1: Check main agent's model configuration
# View main agent's model configuration
cat ~/.openclaw/agents/main/agent/models.json
# View global model providers
cat ~/.openclaw/openclaw.json | grep -A 20 "providers"
Step 2: Choose configuration approach
Option A: Use same model as main agent (recommended)
- New agent can directly use main agent's model configuration
- No additional configuration needed
- Models are shared across agents by default
Option B: Configure new model provider
- Required only if agent needs different model provider
- Add provider to openclaw.json models.providers
- Configure auth-profiles.json for the agent
Step 3: Configure model provider (if needed)
# Add provider to openclaw.json
# Location: ~/.openclaw/openclaw.json
# Example structure:
{
"models": {
"providers": {
"custom-provider": {
"baseUrl": "https://api.example.com/v1",
"apiKey": "key_id:secret",
"api": "openai-completions",
"models": [...]
}
}
}
}
Step 4: Configure agent-specific auth (if needed)
# Edit auth-profiles.json for the agent
# Location: ~/.openclaw/agents/<agent-name>/agent/auth-profiles.json
# Example structure:
{
"custom-provider": {
"apiKey": "key_id:secret",
"baseUrl": "https://api.example.com/v1"
}
}
⚠️ IMPORTANT: Model providers vs Channel providers
-
Model providers: Configure AI model API keys (OpenAI, Anthropic, custom providers)
- Location: openclaw.json → models.providers
- Used for: AI model access
- Examples: openai, anthropic, custom-maas-api-*
-
Channel providers: Configure messaging platform credentials (Feishu, Telegram, Discord)
- Location: openclaw.json → channels.<provider>
- Used for: Message delivery
- Examples: feishu, telegram, discord, whatsapp
Do not confuse these two types of providers!
Common error:
⚠️ Agent failed before reply: No API key found for provider "provider-name"
Solution:
- Check if provider name matches openclaw.json models.providers
- Verify API key is configured correctly
- Check auth-profiles.json for agent-specific configuration
- Restart OpenClaw Gateway after configuration changes
⚠️ IMPORTANT: Never reuse credentials from main agent!
Error recovery:
- If
openclaw agents addfails: Restores from backup and checks syntax - If auth fails: Verify credentials in auth-profiles.json
- If binding fails: Check channel permissions and network connectivity
- If backup fails: Check write permissions on ~/.openclaw/ directory
- If main agent stops responding: Immediately restore from backup and restart OpenClaw
Common errors and solutions:
Error 1: "No API key found for provider 'provider-name'"
Causes:
- Provider name doesn't match any configured provider in openclaw.json
- Typo in provider name
- Provider not added to models.providers
Solutions:
- Check provider name in openclaw.json:
cat ~/.openclaw/openclaw.json | grep -A 10 "providers" - Verify provider name spelling (case-sensitive)
- Add provider to models.providers if needed
- Restart OpenClaw Gateway after configuration changes
Error 2: "Agent failed before reply"
Causes:
- Agent configuration is invalid
- Workspace files are missing or corrupted
- Model configuration is incorrect
Solutions:
- Verify workspace files exist:
ls -la /path/to/workspace/ - Check agent configuration:
openclaw agents list - Test agent loading:
openclaw agents test <agent-name> - Check OpenClaw logs:
openclaw logs --follow
Error 3: "Missing required parameter: path (path or file_path)"
Causes:
- Using wrong parameter name in write tool
- Parameter format error in tool call
Solutions:
- Use correct parameter name:
file_pathinstead ofpath - Verify tool call syntax:
write: file_path: /path/to/file.md content: "content here" - Check tool documentation for correct parameters
Error 4: "Could not find exact text in file"
Causes:
- Using edit tool with incorrect oldText
- Whitespace differences (spaces vs tabs)
- File already modified by another process
Solutions:
- Re-read file to get exact content:
read /path/to/file.md - Use unique markers for large sections:
<!-- SECTION_START --> [content] <!-- SECTION_END --> - Match whitespace exactly (copy from file read)
- Use smaller oldText for more precise matching
Phase 2.6 - Verify configuration
Verification steps:
# 1. Check agent is registered
openclaw agents list
# 2. Verify workspace files exist
ls -la /path/to/workspace/
# 3. Test agent can load (dry run)
openclaw agents test <agent-name>
Success criteria:
- Agent appears in
openclaw agents listoutpu
Truncated for display — read the full file on GitHub.
Related Skills
taste-skill
116.8kAgent skills that give coding agents taste: art direction, reference images and a browser check instead of generic AI pages
Claude-Code-Game-Studios
39.7kTurn Claude Code into a full game dev studio — 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy.
x402
121.0kSet up Browser Use Cloud payments with x402 — pay per request from a crypto wallet (USDC on Base mainnet), no signup or API key.
algorithmic-art
177.9kCreating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
