Code Execution Guide

Overview

This document serves as the definitive operational manual for AI agents acting within the kg-automation repository. It dictates the agent's role, required startup procedures, and architectural constraints to ensure reliable system execution.

Agents using these instructions must treat themselves as the 'execution layer,' meaning they implement designs rather than creating them. Adherence to this guide is mandatory for maintaining system integrity.

Capabilities

• Mandatory Startup Sequence: Agents must begin every session by reading CLAUDE.md, checking the design roadmap, reviewing recent git logs, and confirming a clean working tree.
• Platform Awareness: The agent must be aware of the target environment (Linux/office2) and avoid assuming functionality on unsupported platforms (like Windows).
• Service Interaction: It details key services available on office2, including OpenClaw, Vikunja, Obsidian Sync, and Restic, specifying how they are accessed.
• Skill Migration Protocol: Provides specific instructions for migrating existing skills, such as adding the Vikunja API task bridge to the inbox-processor without altering core logic.

Example Use Cases

1. Feature Implementation: When tasked with building a new feature, the agent must first consult docs/design/felix-capability-roadmap.md to ensure the design intent is documented before writing any code.
2. Debugging Workflow: If an error occurs, the agent should reference the last 10 commits (git log) and the current status (git status) before attempting a fix.
3. Skill Update: To update the inbox-processor, the agent must follow the precise migration steps outlined, ensuring only the API bridge is added.