What problem does it solve? Coordinating cross-device code changes between a development repository and a build workspace is error-prone: batches must be validated against the current HEAD, edits applied to the right files, verification run on real hardware, and results pushed with receipts. This Skill automates that entire apply-side pipeline. ## Core Features & Use Cases - Batch parsing and gating: Parses CDP batches with base-commit rejection (exit 18 on mismatch), branch and clean-tree prechecks, and budget enforcement. - Structured editing with validation: Applies full-file edits for new files and hunk-level edits for diff files, validated by cdp_validate_patch.py, with AI self-healing retries (up to 3 attempts). - Verification and push orchestration: Triggers workspace-verify via loop-engineering for board testing, writes pass/fail/skip receipts with per-direction self-reports, and pushes everything through git-works-push. - Use Case: A developer pastes a CDP batch from the emit side; the Skill validates the base commit, edits code/rpi5 kernel and AOSP files, regenerates the manifest, runs on-device verification, and pushes the commit with timing receipts. ## Quick Start Paste a CDP batch from the emit side and ask the AI to apply it to the code repository, verify on the workspace, and push the results.
