Calendar Extractor

Periodically scan recent recording sessions, extract calendar events from transcripts, and push a daily summary to your iOS chat.

When to use

• "today's meetings"
• "calendar extract"
• "今日会议"
• "提取日历"

Core commands

# Register (first use)
node scripts/register.js \x3CuserId> \x3Cname>

# Run today's flow (also what the cron triggers)
node scripts/calendar-extractor.js \x3CuserId>

# Push management
node scripts/push-toggle.js on \x3CuserId> [--time HH:MM] [--channel iOS|Telegram|Discord|Slack]
node scripts/push-toggle.js off \x3CuserId>
node scripts/push-toggle.js status \x3CuserId>

Workflow

1. Fetch recent transcripts via the LLM's javis_mcp tools (get_transcript_tool / search_transcripts_tool) and pass the relevant text to this script on stdin.
2. Fetch from the configured HTTP endpoint (set HTTP_SOURCE_URL env var when registering the cron).
3. Format output and POST to http://javis-server:8000/api/agent/push with {"skill": "calendar-extractor", "content": "\x3Cformatted>"} using OPENCLAW_GATEWAY_TOKEN for auth.

Push setup (cron registration)

When user requests scheduled push:

Step 1: Save preferences

node scripts/push-toggle.js on \x3CuserId> --time \x3CHH:MM> --channel \x3Cchannel>

Step 2: Create cron job via openclaw CLI

openclaw cron add \
  --name "calendar-extractor-\x3CuserId>" \
  --schedule "0 8,18 * * *" \
  --tz "America/Los_Angeles" \
  --channel \x3Cchannel> \
  --to "\x3Cchannel-target-id>" \
  --session isolated \
  --command "Run /calendar-extractor: execute node scripts/calendar-extractor.js \x3CuserId>, format output nicely. Then POST to http://javis-server:8000/api/agent/push with JSON body {\"skill\": \"calendar-extractor\", \"content\": \"\x3Cformatted output>\"} using the gateway bearer token for auth."

Step 3: Confirm to user

Push is set up; results land in iOS agent chat under /calendar-extractor.

Supported channels: iOS, Telegram, Discord, Slack

Notes

• Data stored in data/users/\x3CuserId>.json; external HTTP source configured separately via HTTP_SOURCE_URL env var.
• Run npm install before first run (none needed beyond Node 18+ built-ins, but the engines field gates the runtime).
• User IDs only allow letters, digits, -, _ (path-traversal guard in data.js).
• Additional data sources picked but NOT auto-wired in main(): pure-local-state and user-typed-text. The template only emits code for the first 2 sources (transcripts + external-http). To activate the others, hand-edit scripts/calendar-extractor.js:
  • For pure-local-state: const userState = fs.existsSync(safeUserPath(userId)) ? readJson(safeUserPath(userId)) : {};
  • For user-typed-text: const text = process.argv.slice(3).join(' ');
• TZ caveat: cron tz is hardcoded to America/Los_Angeles at registration time. If you travel, re-register with the new tz: node scripts/push-toggle.js off \x3CuserId>, then node scripts/push-toggle.js on \x3CuserId> --tz \x3Cnew-tz>, then re-run the openclaw cron add command above with the new --tz.
• Multi-channel cron: each non-iOS channel (Telegram, Discord, Slack) needs a SEPARATE openclaw cron add with its own --channel and --to \x3Cchannel-target-id>. iOS push is automatic via the /api/agent/push call (no separate cron needed for iOS).
• Session-finalize trigger (NOT implemented): the "fire on each finished recording" requirement is out-of-scope for skill-creator's periodic-push template. It requires server-side wiring: javis-server watches audio_recordings.session_completed flips and POSTs a synthetic prompt to the user's openclaw container /v1/responses to invoke this script with the finalized session id. Track as a follow-up in javis-server.