OpenClaw HowTo

Provides authoritative guidance on OpenClaw features, configs, CLI commands, troubleshooting, using both web search and local info with auto-cached environme...

Security Scan

The skill's files and runtime instructions are consistent with a “how-to” / troubleshooting helper for OpenClaw; it reads/writes local cache files and optionally uses a configured web_search tool, which matches its purpose—no signs of outright malice—but there are a few minor mismatches and privacy/consent points to review before installing.

Purpose & Capability

The name/description (OpenClaw HowTo) aligns with the content: documentation, command cheatsheets, search strategies, and CLI test scripts. The SKILL.md expects to use the openclaw CLI, local reads/writes, and optional web_search tooling—these are reasonable for an on-system help/knowledge skill.

Instruction Scope

Instructions explicitly read and write local cache/knowledge files (memory/openclaw-env-info.md, memory/openclaw-knowledge/...), run openclaw CLI commands, and may invoke an external web_search via mcporter if available. This is consistent with the stated dual-mode behavior, but the skill auto-saves configuration after first use (persistent cache) which has privacy/consent implications; the instructions should be reviewed if you do not want automatic local persistence.

Install Mechanism

There is no install spec (instruction-only) and a single harmless test script. Nothing is downloaded or extracted from remote URLs. This is low-risk from an install-mechanism perspective.

Credentials

The skill does not request environment variables or credentials. It uses local CLI tools (openclaw, optionally mcporter) and the agent's read/exec capabilities, which are proportional to a troubleshooting/how-to skill. There are no hard-coded external secrets or endpoints in the files.

Persistence & Privilege

The skill persists a cache file (memory/openclaw-env-info.md) and writes knowledge artifacts under memory/. It is not set always:true and doesn't modify other skills. Still, persistent caching occurs after first use and the SKILL.md states it 'will remember user configuration after first use and no longer ask repeatedly'—confirm you consent to that behavior and review what is stored.

Assessment

This skill appears to be what it says: a local how-to that runs openclaw commands, stores local caches, and optionally uses a configured web_search tool via mcporter. Before installing or using it: 1) Confirm you are comfortable with the skill writing files under memory/ (memory/openclaw-env-info.md and memory/openclaw-knowledge/). 2) Do not paste API keys or secrets when providing a web_search configuration; the skill will save whatever you provide. 3) Note a minor metadata mismatch: the SKILL.md expects the openclaw CLI (and may call mcporter if present) while the registry metadata listed no required binaries—verify your environment has openclaw and (optionally) mcporter if you want web-search features. 4) You can inspect or delete the memory files if you want to revoke persisted config. If you want stricter privacy, run the skill in a sandboxed workspace or deny its ability to write to memory/ before use.

OpenClaw HowTo - OpenClaw HowTo Skill

Use this skill to become an OpenClaw expert, accurately understanding OpenClaw's capabilities, configurations, environment requirements, and the latest information to provide users with the most intuitive and effective answers.

🔍 Intelligent Dual-Mode Search Strategy

Core Mechanism: Auto Memory + Graceful Degradation

• ✅ One-time Configuration, Permanent Recall → Saved to memory/openclaw-env-info.md
• ⏰ 7-Day Validity → Prompt re-validation when expired
• 🔄 Cache Priority → No repeated questions
• 🛡️ Automatic Fallback → Switches to local mode when web_search is unavailable

---

🚀 Step 1: Runtime Detection Flow

def get_websearch_tool():
    """
    Logic for obtaining web_search tool:
    1. First check cache (memory/openclaw-env-info.md)
    2. If cache invalid/missing, detect current environment
    3. If new tool detected, save to cache
    4. If none available, return None (fallback to local mode)
    """
    
    # Step 1: Check cache
    cached = read_cached_config("memory/openclaw-env-info.md")
    if cached and is_cache_valid(cached, days=7):
        return {"source": "cached", **cached}
    
    # Step 2: Attempt environment detection
    env_tools = detect_env_tools()
    if env_tools:
        save_new_config(env_tools)
        return {"source": "auto-detected", **env_tools}
    
    # Step 3: Only prompt on first use with no available tools
    if not has_asked_before():
        prompt_user_for_config()
        return None  # Let user decide whether to continue
    
    # Step 4: Fallback to local mode
    return None

---

📁 Cache File Format

memory/openclaw-env-info.md (auto-maintained):

## Environment Info
- **Detection Time**: 2026-03-28 23:59
- **OpenClaw Version**: 2026.3.8

### Web Search Tool Configuration (Cached)
- **Source**: auto-detected / user-provided / cached
- **Tool Name**: MiniMax.web_search
- **Call Format**: mcporter call [tool_name] query:"[query]"
- **Success Example**: mcporter call MiniMax.web_search query:"OpenClaw cron help"
- **Saved At**: 2026-03-28 23:59
- **Last Verified**: 2026-03-28 23:59
- **Status**: ✅ Active / ❌ Disabled

---

## How to Configure/Update web_search Tool

```bash
# Run the following command to start configuration
openclaw configure web-search --setup

---

Note:

• Config file will be auto-saved as skill updates
• Will prompt re-test if not verified for more than 7 days
• Deleting this file will reset to first-time state

---

## 🔄 Complete Workflow

### Scenario A: Cache Exists (After First Use)

User Question → Read Cache → Validate Validity (<7 days) ↓ ↓ ↓ Use Cached Tool ← ✅ Valid → Use Directly ↓ ❌ Expired → Prompt Re-verification

**Output Example**:

```markdown
💡 Using configured web_search tool to search...

Tool: MiniMax.web_search  
Last Verified: Yesterday

Searching: "OpenClaw Cron Job Configuration"

---

Scenario B: First Time (No Cache)

User Question → Detect Environment → Found No web_search
   ↓               ↓           ↓
First Prompt   ←      ❌         ✅ Has Tool
                       ↓       ↓
                    Ask User → Wait for Choice
                     ↓       ↓
                  Option A  Option B
                   ↓       ↓
                Collect Config  Direct Fallback
                   ↓       ↓
                Save to Memory

Output Example:

⚠️ No available web_search tool detected

To obtain the latest information, you can choose:

**Option A: Provide Existing Configuration**  
If you have other web_search tools, please share the following:
- Tool Name (e.g., "MiniMax.web_search")
- Call Method (e.g., "mcporter call [name] query:'[text]'")
- A Success Example

**Option B: Use Local Mode**  
Skip web_search, answer only via CLI and existing knowledge base

Please choose A or B 👇

---

Scenario C: After User Provides Configuration

Receive User Input → Parse Config → Verify Availability → Save to Memory
   ↓              ↓         ↓          ↓
Format Validation  Extract Fields    Test Once  memory/openclaw-env-info.md
   ↓              ↓         ↓          ↓
❌ Invalid Format → Prompt Correction  ✅ Success  Record Timestamp
                            ↓
                      Confirm Save and Continue

Output Example:

✅ Configuration successfully remembered!

📝 **Saved Information:**
- Tool Name: MiniMax.web_search
- Call Format: mcporter call [tool_name] query:"[query]"
- Status: ✅ Active
- Saved At: 2026-03-28 23:59

💡 Will load automatically next time, no need to ask again.

Now continuing with your request...

---

💬 Interaction Templates

1. Prompt When Using Cache

💡 **Web Search Mode** (Configured)

Using saved web_search tool to find latest information:

**Tool**: `MiniMax.web_search`  
**Last Verified**: Today  
**Search Scope**: OpenClaw Official Docs + GitHub + Discord

Please wait...

---

2. Prompt When Configuration Expires

⚠️ **Web Search Configuration Needs Re-verification**

Last verification was `X` days ago (>7 days), to ensure accuracy:

(A) Re-verify tool configuration  
(B) Temporarily fallback to local mode

Or do you want to update to a different tool?

---

3. Fallback to Local Mode

⚠️ **Currently in Local Mode** (web_search unavailable)

Will provide assistance through:
- Reading OpenClaw CLI commands
- Analyzing configuration files
- Consulting local knowledge base (`memory/openclaw-knowledge.md`)

💡 **Hint**: For latest information, configure web_search tool and try again.

Here is the answer based on existing knowledge:

---

🛠️ Automation Script Recommendations

save-websearch-config.py

#!/usr/bin/env python3
"""Auto-save web_search config to memory/openclaw-env-info.md"""

from datetime import datetime

def save_config(tool_name, call_format, example=None):
    config = {
        "tool_name": tool_name,
        "call_format": call_format,
        "example": example or f"mcporter call {tool_name} query:'test'",
        "saved_at": datetime.now().isoformat(),
        "last_verified": datetime.now().isoformat(),
        "status": "active"
    }
    
    with open("memory/openclaw-env-info.md", "a") as f:
        f.write("\n### Web Search Tool Configuration\n")
        for k, v in config.items():
            f.write(f"- **{k}**: {v}\n")
    
    print("✅ Configuration saved to memory/openclaw-env-info.md")

---

validate-websearch.sh

#!/bin/bash
# Verify if cached configuration is valid

ENV_FILE="memory/openclaw-env-info.md"
LAST_VERIFIED=$(grep "Last Verified" $ENV_FILE | awk '{print $3" "$4}')
CURRENT_DATE=$(date +"%Y-%m-%d %H:%M")

DAYS_OLD=$(( ($(date -d "$CURRENT_DATE" +%s) - $(date -d "$LAST_VERIFIED" +%s)) / 86400 ))

if [ $DAYS_OLD -gt 7 ]; then
    echo "⚠️ Web Search configuration expired ($DAYS_OLD days ago)"
    exit 1
else
    echo "✅ Web Search configuration valid (${DAYS_OLD}days ago verified)"
    exit 0
fi

---

🎯 Trigger Scenarios

Use this skill when users ask about:

• "Does openclaw support..."
• "Help me use openclaw to do/find/accomplish..."
• "Let me see how openclaw does..."
• "How to configure openclaw..."
• "How to set up openclaw cron jobs"
• "How to create a new agent"
• "How to manage openclaw skills"
• Any implied questions about OpenClaw-related information

---

📚 Knowledge Base Structure

Maintain the following files under memory/ directory:

memory/openclaw-knowledge.md

OpenClaw Core Knowledge Base (command reference, configuration options, best practices)

memory/openclaw-env-info.md (New)

Environment Info Recording (web_search tool config, version info, special configs)

memory/openclaw-commands.md

OpenClaw Command Quick Reference

---

📖 Common Command Reference

Basic Commands

# View OpenClaw status
openclaw status

# View all available commands
openclaw help

# View help for specific command
openclaw [command] --help

Agent Management

# List all agents
openclaw config get agents.list --json

# View agent details
openclaw config get agents.list | jq '.[] | select(.id == "agent-name")'

Skills Management

# List all skills
openclaw skills list

# Install skill
openclaw skills install [skill-name]

# Publish skill
cd ~/my-skill
clawhub publish . \
  --slug my-skill \
  --name "My Skill" \
  --version 1.0.0 \
  --changelog "Initial release"

---

🔍 Troubleshooting

Issue	Solution
Asking about web_search every time	Check if memory/openclaw-env-info.md is saved correctly
Cache config invalid	Manually delete this file to restart configuration
web_search tool call fails	Run mcporter list to check if tool is in list
Network timeout	Check network connection, configure HTTP_PROXY

---

🔄 Configuration Management Commands

# View current web_search config
cat memory/openclaw-env-info.md | grep -A10 "Web Search"

# Clear cache (reconfigure)
rm memory/openclaw-env-info.md

# Re-run skill trigger config flow
openclaw run openclaw-howto --setup

---

Last Updated: 2026-03-28
Maintainer: OpenClaw HowTo Skill
Goal: One-time configuration, permanent recall, intelligent fallback
Version: 2.1 (With Cached Memory)