codex_playwright_skills_staging

This repo contains staged Codex skills for working with the Playwright MCP Chrome bridge in a very particular setup.

This is a staged development repo, not the live installed CODEX_HOME skill directory.

Why this exists

Using Codex with the Chrome Playwright MCP bridge turned out to require real investigation. The workflow is not self-evident, and several parts of the connection model had to be discovered, tested, and written down carefully before the setup became reliably usable.

These skills exist to encode that repeatable workflow so the user does not have to keep rediscovering:

• how initial connection works
• how Codex and the user stay synchronized on the bound tab
• what disconnect and recovery look like
• when Playwright is the right intake path for a webpage task

In practice, this skill-guided workflow has noticeably improved usability when working with Codex in this environment.

Scope

This repo is tuned for:

• Codex
• the Playwright MCP browser extension / Chrome bridge workflow
• the specific operational behavior discovered in this environment

It should not be treated as a general Playwright guide.

In another setup, another client, another browser integration, or another MCP environment, parts of this workflow may be unhelpful or misleading.

What is here

• playwright-mcp-webpage-intake
  • guidance for deciding when Playwright MCP is the right way to show Codex a webpage and which connection mode makes sense
• playwright-mcp-connected-tab-workflow
  • the operational workflow for initial selection, bound-tab verification, working inside the bound tab, and recovery after disconnect

Relation to installed skills

These folders are staged development copies of the skills.

They are used to:

• develop and revise the skills in a controlled place
• preserve branch and tag history across revisions
• validate and test candidate versions before installation

They are not the same thing as the live installed skill copies under CODEX_HOME.

The installed skills are deployment targets. This repo is the versioned staging surface used to develop and test changes before promoting them into the installed environment.

That distinction matters:

• this repo preserves development history intentionally
• the installed skills may lag behind or differ temporarily depending on what has or has not been deployed yet
• no one should assume this repo is automatically identical to the current installed CODEX_HOME version

Why the staged repo was published

This repo was published from staging rather than from the installed skill directory on purpose.

What staging preserves better:

• the full development history from v1 to v2 to v3
• a cleaner and more focused repository without unrelated local installation concerns
• a safer place to iterate, validate, and test before deployment

What the installed directory is better for:

• live runtime use by Codex
• deployment, not development history

So the GitHub repo should be understood as the curated development and testing history of these skills, not as a claim that it is always the exact live installed state.

Development history

This repo keeps the staged development history on purpose.

• v1 is the imported baseline
• v2 captures the tokenless tab-selection and session-recovery update
• v3 refines verification around browser_snapshot and the supporting role of browser_tabs

The branch and tag history is meant to preserve how the workflow was discovered and refined, not just the latest version.