L76 Core Architecture Skill
This skill demonstrates a complete, production-ready AgentSkills architecture.
When to Use
✅ USE this skill when:
- Building a new skill from scratch
- Reviewing or auditing existing skill structure
- Learning AgentSkills best practices
- Needing a template for tool integration
Architecture Overview
l76-core-arch/
├── SKILL.md # Skill manifest + instructions (YOU ARE HERE)
├── index.js # Main entry point (optional for simple skills)
├── references/ # Supporting docs, examples, templates
│ └── examples.md
└── scripts/ # Helper scripts (optional)
└── validate.sh
Core Components
1. Frontmatter (Required)
---
name: skill-name # kebab-case, unique identifier
description: | # Clear trigger conditions
When to use this skill.
metadata:
author: your-name
version: "1.0.0"
emoji: "🎯" # Optional visual identifier
openclaw: # OpenClaw-specific metadata
requires:
bins: ["tool1", "tool2"]
install: # Installation instructions
- id: tool1
kind: node
package: package-name
bins: ["binary-name"]
label: "Install description"
---
2. Skill Instructions Structure
# Skill Name
Brief description (1-2 sentences).
## When to Use
✅ **USE this skill when:**
- Clear trigger condition 1
- Clear trigger condition 2
❌ **DON'T use this skill when:**
- Anti-pattern 1
- Anti-pattern 2
## Workflow
### Step 1 — Preflight
Check prerequisites before starting.
### Step 2 — Main Logic
Core skill functionality.
### Step 3 — Cleanup
Finalize and report results.
## Error Handling
Common errors and recovery steps.
## Examples
Concrete usage examples.
Tool Integration Patterns
Pattern 1: Sequential Tool Calls
// Read file, then process, then write
const content = await read({ path: "input.txt" });
const processed = transform(content);
await write({ path: "output.txt", content: processed });
Pattern 2: Conditional Execution
// Check first, then act
const exists = await exec({ command: "test -f file.txt" });
if (exists.exitCode === 0) {
await edit({ path: "file.txt", oldText: "...", newText: "..." });
}
Pattern 3: Error Recovery
try {
await riskyOperation();
} catch (error) {
// Log error
await write({ path: "error.log", content: error.message });
// Attempt recovery
await recoveryStep();
// Report to user
return { status: "recovered", error: error.message };
}
Pattern 4: Batch Operations
// Process multiple items efficiently
const items = await discoverItems();
const results = await Promise.all(
items.map(item => processItem(item))
);
Error Handling Strategy
Error Categories
- Recoverable — Retry with backoff, fallback to alternative
- User Action Required — Prompt user for input/permission
- Fatal — Report clearly, suggest workaround
Error Response Format
{
status: "error" | "partial" | "success",
error: "Human-readable error message",
recovery: "Suggested next step",
details: { /* Technical details */ }
}
Memory Items
Skills should track state when needed:
### Skill State
- Last run: 2026-03-22
- Items processed: 42
- Errors encountered: 3
- Configuration: {...}
Testing Checklist
Publishing to ClawHub
# Login first
clawhub login
# Publish
clawhub publish ./l76-core-arch \
--slug l76-core-arch \
--name "L76 Core Architecture" \
--version 1.0.0 \
--changelog "Initial release"
# Verify
clawhub list
References
