Openclaw Automation Recommender

OpenClaw Automation Recommender

Analyze workspace patterns to recommend tailored OpenClaw automations across all extensibility options.

This skill is read-only. It analyzes the workspace and outputs recommendations. It does NOT create or modify any files. Users implement the recommendations themselves or ask an agent separately to help build them.

Output Guidelines

• Recommend 1-2 of each type: Don't overwhelm — surface the top 1-2 most valuable automations per category
• If user asks for a specific type: Focus only on that type and provide more options (3-5 recommendations)
• Go beyond the reference lists: The reference files contain common patterns, but use web search to find recommendations specific to the workspace's tools, frameworks, and libraries
• Tell users they can ask for more: End by noting they can request more recommendations for any specific category

Automation Types Overview

Workflow

Phase 1: Workspace Analysis

Gather project context:

# Check OpenClaw configuration
ls -la ~/.openclaw/openclaw.json
cat ~/.openclaw/openclaw.json 2>/dev/null | head -100

# Check existing cron jobs
openclaw cron list 2>/dev/null

# Check existing skills
ls -la ~/.openclaw/skills/ 2>/dev/null

# Check workspace structure
ls -la ~/.openclaw/workspace/ 2>/dev/null

# Check agent workspaces
ls ~/.openclaw/workspace/agents-workspaces/ 2>/dev/null

# Check processes
cat ~/.openclaw/workspace/processes/PROCESSES.json 2>/dev/null | head -50

# Check project files
ls ~/.openclaw/workspace/projects/ 2>/dev/null

# Analyze project structure if applicable
ls -la src/ app/ lib/ tests/ components/ pages/ api/ 2>/dev/null

Key Indicators to Capture:

Phase 2: Generate Recommendations

Based on analysis, generate recommendations across all categories:

A. Cron Job Recommendations

Cron jobs are scheduled via openclaw cron and run with --no-deliver --session isolated.

Cron job best practices:

• Always use --no-deliver (critical — even best-effort with no chat window records error)
• Always use --session isolated (sandboxed, fresh context per run)
• Use --model override for specific providers (e.g., --model deepseek/deepseek-chat for unlimited)
• Never dispatch Anthropic-primary agents concurrently (shared OAuth rate limits)
• Max 1 dispatch per cron run (prevents race conditions on tickets.json)

B. Skills Recommendations

See references/skills-reference.md for details.

Create skills in ~/.openclaw/skills/<name>/SKILL.md:

C. Agent Routing Recommendations

OpenClaw has a 16-agent roster with MoE (Mixture of Experts) tag-based dispatch:

Agent routing best practices:

• Cooper (main) orchestrates — routes, delegates, synthesizes
• Nova is R&D-only — never assign regular tickets
• Pixel uses Opus (expensive) — reserve for complex debugging
• MoE routing is tag-based, not availability-based
• Each agent has idle-time standing orders when queue + backlog empty

D. Process Recommendations

Processes are formalized workflows in ~/.openclaw/workspace/processes/PROCESSES.json with maturity levels:

Process maturity levels:

• L1: Ad hoc — just documented
• L2: Documented — has steps and roles
• L3: Practiced — used 3+ times with evidence
• L4: Optimized — measured and improved

E. MCP Server Recommendations

See references/mcp-servers.md for detailed patterns.

F. Workspace Structure Recommendations

Phase 3: Output Recommendations Report

Format recommendations clearly. Only include 1-2 recommendations per category — the most valuable ones. Skip categories that aren't relevant.

## OpenClaw Automation Recommendations

I've analyzed your workspace and identified the top automations for each category.

### Workspace Profile
- **Agents Active**: [count]/16
- **Cron Jobs**: [count]/17
- **Skills**: [count]
- **Processes**: [count]/23
- **Projects**: [count]

---

### ⏰ Cron Jobs

#### [cron name]
**Why**: [specific reason based on detected gaps]
**Setup**: `openclaw cron add --name "[name]" --schedule "[cron]" --agent [agent-id] --no-deliver --session isolated`
**Model Override**: `--model [provider/model]`

---

### 🎯 Skills

#### [skill name]
**Why**: [specific reason]
**Create**: `~/.openclaw/skills/[name]/SKILL.md`

---

### 🧠 Agent Routing

#### [routing recommendation]
**Why**: [specific reason based on workspace patterns]
**Tag**: [dispatch tag]
**Agent**: [agent name] ([agent id])

---

### 📋 Processes

#### [process name]
**Why**: [specific reason]
**Maturity Target**: L[level]
**Add to**: `workspace/processes/PROCESSES.json`

---

### 🔌 MCP Servers

#### [server name]
**Why**: [specific reason based on detected services]
**Install**: [install command]

---

### 📁 Workspace Structure

#### [structure recommendation]
**Why**: [specific reason]
**Create**: [paths to create]

---

**Want more?** Ask for additional recommendations for any specific category.

**Want help implementing any of these?** Just ask and the appropriate agent can help set up any of the recommendations above.

Decision Framework

When to Recommend Cron Jobs

• Recurring tasks that should happen on schedule
• Monitoring and health checks
• Synthesis and reporting
• Memory and knowledge maintenance
• R&D continuous exploration

When to Recommend Skills

• Frequently repeated prompts or workflows
• Project-specific tasks with templates
• Domain expertise that agents should apply automatically
• Quick actions the user invokes regularly

When to Recommend Agent Routing Changes

• Tasks consistently going to the wrong agent
• Agents overloaded while others are idle
• New domains or specializations needed
• Quality issues traceable to agent mismatch

When to Recommend Processes

• Workflows that happen repeatedly without structure
• Quality issues from inconsistent approaches
• Team coordination breakdowns
• Compliance or audit requirements

When to Recommend MCP Servers

• External service integration needed (databases, APIs)
• Documentation lookup for libraries/SDKs
• Browser automation or testing
• Team tool integration (GitHub, Linear, Slack)

When to Recommend Workspace Structure

• Files scattered without organization
• Missing communication channels between agents
• No project or sprint tracking
• Knowledge not being captured

Configuration Tips

Cron Job Setup

# Add a new cron job
openclaw cron add --name "my-cron" --schedule "0 */6 * * *" --agent cipher --no-deliver --session isolated

# List existing cron jobs
openclaw cron list

# Remove a cron job
openclaw cron remove --name "my-cron"

Agent Dispatch

Cooper handles routing via MoE tags in CLAUDE.md. To update routing:

1. Edit ~/.openclaw/workspace/CLAUDE.md routing table
2. Update per-agent SOUL.md if role changes
3. Verify with openclaw doctor

Provider-Aware Scheduling

When scheduling cron jobs, consider provider rate limits:

• Anthropic agents (Cooper, Pixel, Vault): Stagger by 120s minimum, share OAuth limits
• DeepSeek agents (Oracle, Nova, Mirror, Forge, Cipher): No hard limits, safe to overlap
• MiniMax agents (Sage, Vista, Anchor, Muse, Vex): 500 RPM, generous limits
• NIM agents (Axon, Vigil): 40 RPM free tier, use sparingly

Headless Mode (for CI/Automation)

# Run a one-off task
openclaw run --agent forge --task "fix lint errors in src/" --no-deliver

# Scheduled task with model override
openclaw cron add --name "nightly-lint" --schedule "0 3 * * *" --agent forge --model deepseek/deepseek-chat --no-deliver --session isolated