
from openclaw-rpa196
Record browser, Excel, Word and API actions once and replay them as standalone Playwright Python scripts—fast, cheap, and deterministic.
OpenClaw RPA provides an LLM-assisted record-and-replay framework that captures browser, API, Excel and Word workflows and exports them as standalone Playwright Python scripts. Record a flow once with screenshots and verification; replay the generated script later with no model calls, avoiding recurring LLM costs and hallucinations.
Use this skill when you need repeatable automation for web scraping, form-filling, API calls, Excel/Word report generation, or cross-step workflows (e.g., API → Excel → Word). It's ideal for tasks that benefit from deterministic replays, cost savings, and proof-of-run screenshots. Not recommended for heavy OS automation or flows that require interactive 2FA.
rpa/ scripts and a registry.json of recorded tasks (has_scripts=false flag here refers to discovery metadata).#rpa-list / #rpa-run:{name} triggers, config.json locale routing, and output format (Playwright Python).Compatible with agents and toolchains that can call Playwright or run Python scripts (OpenClaw-style assistants, local Python RPA runners).
OpenClaw RPA is an LLM-based browser/file automation framework that records actions and replays them as standalone Playwright Python scripts. Scripts are well-structured with proper error handling; bootstrap_config.py ran cleanly, others failed due to sandbox environment constraints (missing venv package, missing CLI args) rather than code bugs. No security concerns — no hardcoded secrets, no destructive commands, no exfiltration risk. The locale-routing architecture (en-US/zh-CN) and progressive disclosure are well-designed.
python3-venv (needed by install.sh for .venv creation)Solid skill with clear documentation and good separation of concerns. The record-once-replay-many pattern is a genuine cost-saving approach. Locale support for zh-CN/en-US is thoughtful. No security issues found.