| name | ant-chrome-openclaw-automation-plan |
| description | Use when implementing, reviewing, auditing, or updating the ant-chrome and OpenClaw integration design, rollout status, or public HTTP API contract. Prefer this skill for low-coupling integration work around runtime/session, runtime/stop, automation/scripts, playwright-cdp execution, and OpenClaw compatibility mapping. Do not use it for day-to-day LaunchServer API invocation; use the sibling ant-chrome-openclaw skill instead. |
Ant Browser OpenClaw Integration Plan
Use this skill for engineering work on the integration itself: planning, code changes, API alignment, audits, and rollout decisions.
If the task is operational usage of LaunchServer from OpenClaw, read the sibling skill at ../ant-chrome-openclaw/SKILL.md instead.
Read Only What You Need
- Read
references/overview.md for the architecture split, scope, and recommended integration path.
- Read
references/api-contract.md for the exact public API surface that exists today.
- Read
references/delivery-status.md for what is actually shipped versus intentionally deferred.
- Read
references/script-compat.md for how playwright-cdp and script execution should absorb LaunchServer, CDP, and Playwright complexity.
Working Rules
- Keep OpenClaw as the orchestration layer and Ant Browser as the execution layer.
- Prefer public/common APIs first:
POST /api/runtime/session
POST /api/runtime/status
POST /api/runtime/stop
GET /api/automation/scripts
POST /api/automation/scripts/run
GET /api/automation/scripts/runs
- Do not add
/api/automation/openclaw/* private routes unless the user explicitly needs a compatibility alias for an already-fixed external protocol.
- Do not re-invent a second automation runtime for OpenClaw; reuse the existing script execution chain.
- Keep the external script protocol object-shaped (
selector / params) and only convert to internal text fields at the boundary.
- Treat
playwright-cdp as the main complexity-reduction layer for OpenClaw-facing automation.
Validation
After changing code or the plan:
- Re-check the current LaunchServer routes under
backend/internal/launchcode.
- Run
go test ./backend/test/launchcode/....
- Run
go test ./backend/....
- Run
python C:\Users\Administrator\.codex\skills\.system\skill-creator\scripts\quick_validate.py <this-skill-dir>.
- If
SKILL.md changes meaningfully, regenerate agents/openai.yaml.