---
name: openclaw-reporter
description: |
  Opt-in reporter for the OpenClaw global lobster heatmap.
  On first use, ASKS the user for consent before registering.
  Sends heartbeats (platform + model only) and generic task summaries to https://kymr.top/.
  All network calls run synchronously in the foreground — output is visible to the user.
  Data collected: user-chosen lobster name, OS platform, model name, generic task category.
  Credentials: the server returns an apiKey on registration, which is stored locally
  in ~/.openclaw/config.json and sent as a Bearer token in subsequent requests.
  No file paths, code, or project-specific data is ever sent.
requirements:
  network: true
  writeFiles:
    - "~/.openclaw/config.json"
  binaries:
    - curl
    - python3
    - uname
  env:
    - name: CLAUDE_MODEL
      required: false
      description: "Current Claude model identifier. Used to report which model is active. Falls back to 'unknown' if not set."
---

# OpenClaw Reporter

This skill lets you participate in the OpenClaw global lobster activity heatmap. It reports **anonymous, non-identifying** activity data to visualize worldwide AI agent usage.

## Data Disclosure

This skill sends the following data to `https://kymr.top/`:

| Data Field | Example | Purpose |
|---|---|---|
| Lobster Name | `CoolLobster42` (user-chosen) | Your display name on the heatmap |
| Platform | `darwin`, `linux` | OS type for heatmap stats |
| Model | `claude-sonnet-4-6` | Model usage stats |
| Task summary | `"Fixed a bug"` | Generic activity category |

**Stored locally:** Registration returns an `apiKey` from the server, saved in `~/.openclaw/config.json`. This token is sent as `Authorization: Bearer` header in heartbeat and task report requests. The `CLAUDE_MODEL` environment variable is read (if available) to identify the active model — it falls back to `"unknown"` if not set.

**Never sent:** system usernames, file paths, code snippets, project names, secrets, IP-derived geolocation beyond what the server naturally receives.

## Configuration

Config is stored at `~/.openclaw/config.json`. The skill checks for this file at the start of each conversation.

## First-Time Setup (Requires User Consent)

If `~/.openclaw/config.json` does not exist, follow these two steps:

### Step 1: Ask for consent

Use the AskUserQuestion tool:

- Question: "OpenClaw Reporter wants to register a lobster name and send activity heartbeats to https://kymr.top/ for the global heatmap. The server will issue an API key stored locally at ~/.openclaw/config.json. No file paths, code, or project data is sent. Allow?"
- Options: "Yes, register" / "No, skip"

**If the user selects "No, skip":** do NOT register and do NOT ask again in this session. Stop here.

### Step 2: Ask for a lobster name

If the user consented, use AskUserQuestion again:

- Question: "Pick a name for your lobster on the heatmap! (e.g. CoolLobster, NightCrawler, CodeClaw)"
- Let the user type a custom name via the "Other" option (no preset options needed, but provide a few fun suggestions as options)
- Options: "CoolLobster" / "CodeClaw" / "NightCrawler"

Use whatever the user provides as the lobster name.

### Step 3: Register

After receiving the name (stored as `LOBSTER_NAME`):

```bash
# Create config directory
mkdir -p ~/.openclaw

OPENCLAW_ENDPOINT="https://kymr.top"
LOBSTER_NAME="USER_CHOSEN_NAME_HERE"

RESPONSE=$(curl -s -X POST "$OPENCLAW_ENDPOINT/api/v1/register" \
  -H "Content-Type: application/json" \
  -d "{
    \"name\": \"$LOBSTER_NAME\",
    \"platform\": \"$(uname -s | tr '[:upper:]' '[:lower:]')\",
    \"model\": \"$(echo $CLAUDE_MODEL 2>/dev/null || echo 'unknown')\"
  }")

# Save config
echo "$RESPONSE" | python3 -c "
import json, sys
data = json.load(sys.stdin)
config = {
    'lobsterId': data.get('lobsterId', ''),
    'apiKey': data.get('apiKey', ''),
    'name': '$LOBSTER_NAME',
    'endpoint': '$OPENCLAW_ENDPOINT'
}
with open('$HOME/.openclaw/config.json', 'w') as f:
    json.dump(config, f, indent=2)
print('Registered as:', '$LOBSTER_NAME')
"
```

## Heartbeat (On Every Conversation Start)

If the config file exists, send a heartbeat. The request runs **synchronously in the foreground** with a 5-second timeout, so the user sees it happen:

```bash
if [ -f ~/.openclaw/config.json ]; then
  OPENCLAW_CONFIG=$(cat ~/.openclaw/config.json)
  OPENCLAW_KEY=$(echo "$OPENCLAW_CONFIG" | python3 -c "import json,sys; print(json.load(sys.stdin)['apiKey'])")
  OPENCLAW_ENDPOINT=$(echo "$OPENCLAW_CONFIG" | python3 -c "import json,sys; print(json.load(sys.stdin)['endpoint'])")

  curl -s -o /dev/null -w "OpenClaw heartbeat: HTTP %{http_code}\n" \
    --max-time 5 \
    -X POST "$OPENCLAW_ENDPOINT/api/v1/heartbeat" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OPENCLAW_KEY" \
    -d "{
      \"platform\": \"$(uname -s | tr '[:upper:]' '[:lower:]')\",
      \"model\": \"$(echo $CLAUDE_MODEL 2>/dev/null || echo 'unknown')\"
    }"
fi
```

## Task Completion Report

When a significant task is completed, report a **generic summary only**:

```bash
if [ -f ~/.openclaw/config.json ]; then
  OPENCLAW_CONFIG=$(cat ~/.openclaw/config.json)
  OPENCLAW_KEY=$(echo "$OPENCLAW_CONFIG" | python3 -c "import json,sys; print(json.load(sys.stdin)['apiKey'])")
  OPENCLAW_ENDPOINT=$(echo "$OPENCLAW_CONFIG" | python3 -c "import json,sys; print(json.load(sys.stdin)['endpoint'])")

  curl -s -o /dev/null -w "OpenClaw task report: HTTP %{http_code}\n" \
    --max-time 5 \
    -X POST "$OPENCLAW_ENDPOINT/api/v1/task" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OPENCLAW_KEY" \
    -d "{
      \"summary\": \"TASK_SUMMARY_HERE\",
      \"durationMs\": DURATION_MS_HERE,
      \"model\": \"$(echo $CLAUDE_MODEL 2>/dev/null || echo 'unknown')\",
      \"toolsUsed\": [\"LIST_OF_TOOLS\"]
    }"
fi
```

### Task Summary Rules (STRICT)

- NEVER include file paths, variable names, or code snippets
- NEVER include API keys, secrets, or credentials
- NEVER include project names, repository names, or organization names
- NEVER include personal information (names, emails, usernames)
- ONLY use generic categories: "Implemented a feature", "Fixed a bug", "Refactored code", "Updated configuration", "Added tests"
- Keep summaries under 100 characters

## Opt-Out / Unregister

To stop all OpenClaw reporting, delete the config file:

```bash
rm -f ~/.openclaw/config.json
rmdir ~/.openclaw 2>/dev/null
```

After deletion, no heartbeats or task reports will be sent. The skill will ask for consent again if re-enabled.

## Behavior Rules

1. **User consent is REQUIRED** before first registration — never register silently
2. If the user declines registration, respect the decision and do not ask again in the same session
3. All network calls run **synchronously in the foreground** (no `&`) with `--max-time 5` — the user sees every request
4. If any OpenClaw request fails, continue normally — never interrupt the user's workflow
5. Only report genuinely completed tasks, not intermediate steps
6. Never include identifying information in task summaries