openclaw-main-workspace

Main workspace for OpenClaw agent orchestration.

Summary

This workspace serves as the primary orchestration layer for resume-intake delegation. It receives user requests, normalizes inbound facts into a standard dispatch envelope, and routes tasks to the existing workspace-resume-intake business session. It does not perform direct resume parsing, field extraction, or Bitable writes.

The workspace also contains local environment configuration, memory files, and a node-pairing workflow for connecting a local macOS node to the cloud OpenClaw Gateway via SSH local forwarding.

Included Assets

Core workspace files

• AGENTS.md — session startup sequence, memory management, safety and external action guidelines
• SOUL.md — core agent traits, boundaries, and continuity rules
• TOOLS.md — local environment notes (server hosts, SSH keys, service management)

Node pairing scripts (scripts/openclaw-node-pairing/)

• cloud_show_token.sh — prints gateway token and status from cloud server
• cloud_approve_latest_node.sh — lists pending node requests and provides approve command
• mac_pair_to_cloud.sh — initiates SSH local forward and starts node connection (foreground debug mode)
• mac_node_up.sh — one-touch background start of SSH forward and node
• mac_node_down.sh — one-touch shutdown of SSH forward and node
• mac_node_session.sh — interactive foreground session with auto-cleanup on Ctrl+C

Resume-intake orchestration skill (skills/resume-intake-orchestrator/)

• SKILL.md — orchestrator skill definition
• references/ — orchestration notes, session targeting, delegation prompt, migration plan, compatibility doc map, cutover checklist

Documentation (docs/)

• RESUME_INTAKE_ROUTING.md — when to trigger delegation
• RESUME_INTAKE_SESSION_TARGETING.md — how to target existing worker sessions
• RESUME_INTAKE_ORCHESTRATION_NOTES.md — architecture boundary, dispatch envelope purpose, common failure points
• RESUME_INTAKE_DELEGATION_PROMPT.md — delegation content requirements
• RESUME_INTAKE_SKILL_MIGRATION_PLAN.md — low-risk migration path from root docs to skill structure

Memory (memory/)

• Daily notes files (YYYY-MM-DD.md) documenting session history and technical investigations

How to Use

Session startup

At the start of each session:

1. Read SOUL.md
2. Read USER.md
3. Read today's and yesterday's memory/YYYY-MM-DD.md if they exist
4. In direct chats, also read MEMORY.md

Resume-intake delegation flow

When a user sends a resume PDF or ZIP for Feishu intake:

1. Read the inbound user message as the source of truth
2. Extract attachment facts (file_path, source_name, message_id)
3. Infer input mode (single_pdf, zip_batch, analysis_only, unknown)
4. Build a dispatch envelope using the helper scripts under skills/resume-intake-orchestrator/scripts/
5. Target the existing workspace-resume-intake Feishu direct session via sessions_send
6. Wait for the worker to return a formal result, then reply to the user

Use the skill references in skills/resume-intake-orchestrator/references/ for detailed boundary rules, session targeting logic, and delegation content requirements.

Node pairing (cloud Gateway ↔ local macOS node)

Cloud Gateway endpoint: 127.0.0.1:18789 on server 47.119.177.99

One-time setup — upload scripts to cloud:

scp -i /Users/jianfengxu/Downloads/has_jianfeng_key.pem -r \
  ~/.openclaw/workspace/scripts/openclaw-node-pairing \
  [email protected]:/root/

Daily use — background mode (recommended):

bash ~/.openclaw/workspace/scripts/openclaw-node-pairing/mac_node