clawpatch: AI Reviews AI Code via Local Agents — Pipeline with Full Audit Trail
clawpatch orchestrates local coding agents like Codex CLI and Claude Code as subprocesses to review, fix, and revalidate AI-generated code through a six-stage pipeline (init, map, review, fix, revalidate, open-pr) with semantic slicing, evidence-backed findings, read-only sandboxes, and explicit human approval — without ever calling model APIs directly.
