| name | jido-browser |
| description | Builder-oriented guidance for the upstream `jido_browser` package. Use when Codex needs to build browser-backed Jido automations, wrap navigation or extraction flows in actions, turn browser docs into runnable examples, or review `jido_browser` boundaries versus `req_llm` and app-specific scraping code. |
Jido Browser
jido_browser is the upstream Hex package name.
Start Here
Use this skill when the workflow needs a browser, page state, or DOM interaction rather than plain HTTP requests.
Good triggers:
- "Build a Jido agent that logs in and gathers data from a website."
- "Wrap a browser step in a
Jido.Action."
- "Turn the browser package docs into a runnable demo."
- "Review whether this should use browser automation or a simpler HTTP client."
Read references/builder-notes.md before implementing when selectors, navigation state, or safety limits matter.
Primary Workflows
Build browser-backed actions
- Separate navigation, observation, and mutation steps.
- Keep selectors and site assumptions explicit so the action can fail predictably when a page changes.
- Prefer small page operations over giant "do everything" browser actions.
Decide when browser automation is justified
- Use
jido_browser only when the target behavior depends on rendering, authentication state, or DOM interaction.
- Prefer
req_llm, Req, or API clients for stable JSON or HTML endpoints.
- Minimize the amount of browser state that leaks into the rest of the agent runtime.
Turn docs into runnable examples
- Choose one real browsing scenario such as login, extract, click, or form submit.
- Show the browser step inside a broader Jido workflow only after the page-level behavior is clear.
- Add waits, retries, or assertions only when the docs justify them.
Review boundaries
- Keep DOM and navigation logic in
jido_browser.
- Keep prompt or reasoning loops in
jido or jido_ai.
- Keep site-specific product logic in the application layer.
Build Checklist
- Confirm that the target site really requires browser automation.
- Define selectors, navigation transitions, and failure conditions up front.
- Capture only the page data the next step needs.
- Add tests or examples that tolerate timing and rendering variability.
Boundaries
- Do not use this skill for generic web scraping when plain HTTP is enough.
- Do not hide brittle selectors inside unrelated business logic.
- Do not promise support for site behaviors the docs do not cover.
