resume-context（恢复上下文）

resume-context

Ask OpenClaw about your coding sessions and project notes — powered by the resume CLI and cached via Redis.

Usage

In the OpenClaw dashboard or any connected channel, say:

• "Claw give me a briefing on resume" → session briefing with what you worked on
• "Claw what was I working on for [project]?" → same
• "Claw show me my notes for [project]" → project notes
• "Claw what notes do I have on [project]?" → same

How it works

1. OpenClaw receives your message and triggers this skill
2. The skill finds your project directory (searches ~/Documents/Projects/, ~/, ~/projects/)
3. Checks Redis cache — if fresh (< 5 min), returns instantly
4. On cache miss: runs resume show or resume notes in your project directory
5. resume calls your LLM (Anthropic/Ollama) to generate a plain-English briefing
6. Result is cached in Redis and returned to you

Architecture

You → OpenClaw → resume-context skill ↓ Redis GET (cache hit → instant) ↓ (cache miss) resume show / resume notes ↓ LLM briefing generation ↓ Redis SET (5 min TTL) ↓ Response back to you

Step 1 — Classify intent

• "briefing", "working on", "session", "status", "catch me up" → resume show
• "notes", "note" → resume notes
• Ambiguous → run notes first (fast), then show

Step 2 — Resolve project path

Extract project name from the message (e.g. "for resume" → "resume").

Search in order:

1. ~/Documents/Projects/<name>
2. ~/<name>
3. ~/projects/<name>
4. ~/code/<name>

find ~ -maxdepth 4 -type d -name "<project_name>" 2>/dev/null | grep -v node_modules | grep -v ".git" | head -5

Prefer the match that contains a .resume/ subdirectory. If no project named, find the most recently active session:

ls -t ~/.resume/projects/ 2>/dev/null | head -3

Step 3 — Run the bridge

REDIS_URL="$REDIS_URL" node {baseDir}/resume-mcp.js show <project_path>
REDIS_URL="$REDIS_URL" node {baseDir}/resume-mcp.js notes <project_path>

Step 4 — Return output

The bridge returns JSON with an output field — return it as-is. If cached: true, add "(answered from cache)". If empty: "No session data found. Is resume running in that project?"

Requirements

1. Install resume

cargo install --git https://github.com/nickleodoen/resume

Then set up the shell hook so resume captures your commands:

resume init --install-hook
# Open a new terminal (or source ~/.zshrc / ~/.bashrc)

Start a session in your project:

cd ~/your-project && resume

2. Set ANTHROPIC_API_KEY

Add to your shell profile (~/.zshrc or ~/.bashrc):

export ANTHROPIC_API_KEY=sk-ant-...

resume uses this to generate plain-English briefings via Claude.

3. Get a Redis URL (free)

Sign up at Redis Cloud → New Subscription → Free Tier. Copy the connection string from the database dashboard — it looks like: redis://default:password@host:port

4. Node.js 18+

Required to run the bridge script (resume-mcp.js).

Environment variables

Install

# Via ClawHub
openclaw skills install resume-context

# Manual
cp -r resume-context ~/.openclaw/skills/
cd ~/.openclaw/skills/resume-context && npm install

Add to ~/.openclaw/openclaw.json:

{
  "skills": {
    "entries": {
      "resume-context": {
        "enabled": true,
        "env": {
          "REDIS_URL": "redis://your-redis-url:6379"
        }
      }
    }
  }
}