
from openclaw-skill-claude-code16
Run Claude Code asynchronously with thread-safe Telegram/WhatsApp notifications, heartbeats, and an E2E operator validation workflow for long-running code and r
Provides an async runner for Claude Code that launches detached background tasks (nohup) and delivers progress and final results to Telegram threads or WhatsApp groups. It enforces routing validation, heartbeat updates, mid-task notifications, and a launch confirmation gate so runs are reliable and auditable. Useful for deep research, codebase analysis, refactoring, file generation, and complex multi-step automations where the agent should continue work outside the current assistant turn.
Use this skill when a task is long-running (>60s), multi-step, or requires external tooling and file access (e.g., code refactors, research reports, E2E test runs). Use --validate-only to verify routing before launching into production. Avoid for quick one-off questions or real-time interactive tasks.
Compatible with agents and runtimes that can run detached Python scripts and deliver messages to Telegram/WhatsApp (OpenClaw agent CLI, Claude Code child processes).
Async Claude Code task runner with Telegram/WhatsApp notifications, heartbeat monitoring, session resumption, and thread-safe routing. The bundled openclaw_notify.py helper is a clean notification script but requires mandatory CLI args (--group, --message) so it exits with usage error under dry-run. SKILL.md is extremely thorough with detailed E2E testing protocol, routing safeguards, and troubleshooting guides. No security issues found — all network calls are localhost-only and auth token reads follow declared config patterns.
SKILL.md is exceptionally detailed (~4500 words) which is both a strength and weakness — could benefit from splitting detailed reference material into references/ directory for leaner main doc. Thread-safe Telegram routing and billing guard (ANTHROPIC_API_KEY stripping) are well-designed security features.