agent-harness-design
Designing agent harnesses and tool systems — risk taxonomy for tools, permission decisions, draft/commit pattern, structured tool…
Drive and inspect a local web, IDE, or Electron UI with browser or CDP automation and evidence. Use for UI regressions, screenshots, accessibility snapshots, visual diffs, focus, keyboard, scrolling, resize, console, network, performance, or heap checks. Do not use for
$ npx -y skills add AnastasiyaW/codex-claude-code-config --skill control-ui --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/control-uiContext preview
The summary Claude sees to decide when to auto-load this skill.
Drive and inspect a local web, IDE, or Electron UI with browser or CDP automation and evidence. Use for UI regressions, screenshots, accessibility snapshots, visual diffs, focus, keyboard, scrolling, resize, console, network, performance, or heap checks. Do not use for
name: control-ui description: Drive and inspect a local web, IDE, or Electron UI with browser or CDP automation and evidence. Use for UI regressions, screenshots, accessibility snapshots, visual diffs, focus, keyboard, scrolling, resize, console, network, performance, or heap checks. Do not use for backend-only changes or for a screenshot without a behavior claim.
Verify UI behavior at the real browser or Electron boundary. Reuse the project's Playwright, Cypress, browser, or Electron harness before creating a probe. Keep the data local and disposable.
1. Read the project's documented start command and identify the local URL or debug port. 2. Discover existing browser tests and stable app markers. 3. Select the target page by a positive root marker, role, label, or stable `data-*` attribute, not tab order or coordinates. 4. Capture the initial DOM/accessibility snapshot, screenshot, console state, or network baseline relevant to the claim. 5. Perform one structural action: click, type, keypress, drag, scroll, navigate, or resize. 6. Capture the new state and assert the expected change. 7. Clean up the dev server, debug session, temporary profile, and artifacts.
Use `verify-this` for before/after claims. Use raw CDP only when higher-level APIs cannot provide the required CPU, heap, trace, network, or rendering signal. Do not install Playwright just for a one-off probe when an existing browser tool or dependency is available.
Screenshots, traces, network bodies, and heap snapshots may contain private code or user data. Keep them outside public Git unless they are explicitly sanitized and approved. A public report should contain the command, revision, safe metric, and artifact hash or private location, not the payload.
workflow is correct.
select from the latest structure.
keyboard behavior.
differences; record those inputs before interpreting it.
| Symptom | Likely cause | Action | |---|---|---| | Wrong tab or window | Selector relied on tab order | Enumerate pages and choose a positive app marker | | Click changes nothing | Stale locator or wrong state | Capture a fresh snapshot and wait for the state marker | | Screenshot differs only on one machine | Fonts, viewport, scale, or locale drift | Pin the test inputs and classify as `INCONCLUSIVE` until comparable | | Browser remains after the run | Cleanup path missed an exception | Use a bounded cleanup step and verify the process/profile is gone |
Adapted from Cursor Team Kit's MIT-licensed `control-ui` workflow: https://github.com/cursor/plugins/tree/main/cursor-team-kit/skills/control-ui
A practical configuration kit for Claude Code, Codex, and other coding agents. It contains architectural principles, enforcement hooks, skills, drop-in rules, starter templates, and dynamic-workflow commands.
Designing agent harnesses and tool systems — risk taxonomy for tools, permission decisions, draft/commit pattern, structured tool…
Use when designing, auditing, refactoring, or explaining an agentic harness for any domain, especially when work must continue…
Практическая инженерия диффузионных моделей: архитектуры, обучение, инференс, оптимизация памяти. Использовать при любых задачах…
Expert prompt engineering for FLUX.2 [klein] image generation and editing model. Use this skill whenever the user wants to create…
Plan or review LoRA and edit-training work specifically for FLUX.2 Klein or Qwen-Image-Edit, including paired datasets,…
Forensic image-to-prompt compiler for image generation models. Use this skill whenever the user wants to: convert/describe an…