Vibecoding Pro

Transform your AI coding workflow from "write and hope" to "iterate with precision." VibeCoding Pro implements the Generator-Evaluator dual-agent pattern (in...

This package is conceptually coherent for generator/evaluator workflows, but before installing or running anything: 1) Review and run a manual code audit of the two Python scripts — they contain NotImplementedError stubs and example calls (sessions_spawn) you must implement; do not run unreviewed code. 2) Expect to install Playwright and a browser (Chromium/Firefox) and possibly node/python toolchains — the skill does not declare these; install them in an isolated environment. 3) Prepare any deployment and API credentials you'll supply to the Evaluator/Generator and keep them minimal (use short-lived tokens where possible). The skill's templates accept auth headers and may require DB/API access — only provide credentials needed for the specific artifact under test. 4) When adapting run_generator/run_evaluator, avoid embedding secrets in code or returning them in evaluation JSON. 5) If you plan to wire this to a remote agent platform (sessions_spawn examples), verify and audit those platform calls to ensure they don't leak data. If you need this to run autonomously, add explicit declarations of required binaries and env vars and restrict tokens to least privilege. If any of these points are unclear or you want, I can enumerate exactly which env vars and binaries you should add to the manifest for a typical Playwright-based setup.

Type: OpenClaw Skill Name: vibecoding-pro Version: 1.0.0 The 'vibecoding-pro' skill bundle is a well-documented framework for implementing a Generator-Evaluator dual-agent workflow, aimed at improving AI-generated code quality through independent QA. The bundle includes architectural documentation (architecture.md), prompt templates (evaluator-prompts.md), and Python script templates (iteration_loop.py, calibrate_evaluator.py) that use standard OpenClaw patterns like subagent spawning and Playwright integration. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found; the references to 'Anthropic 2026' research appear to be a stylistic or fictional framing for the methodology rather than an attack vector.