ego-jev

# ego-jev ego executes; Jev decides. Each step: snapshot → number the interactive elements (a11y refs `@N` + DOM-discovered clickables `dN`: `div` cards, `cursor:pointer`, iframe/shadow content) → one System One call picks the operation and its target together → ego clicks/fills/selects. Jev never writes text and never sees a screenshot. Read `ego-browser` first for TaskSpace and Page rules; this skill only replaces the per-step "which element" judgment. Requires: [ego lite](https://lite.ego.app/) installed and onboarded (provides the `ego-browser` command), the `ego-browser` skill, and `TYPESAFE_API_KEY` or `AI_GATEWAY_API_KEY` in `~/.config/ego-jev/secrets.env`. ## Run 1. Have a Page on the starting URL (`goto` with `waitUntil: "domcontentloaded"` on slow marketing sites; the default `load` times out at 15 s). 2. Call the loop with the goal verbatim, every value Jev may need to type, and a `verify` that tests the real done state (`SKILL_DIR` is the absolute path of the directory containing this SKILL.md): ```js const { runJevLoop } = await import( `file://${SKILL_DIR}/scripts/jev-loop.mjs` ); const result = await runJevLoop(page, { goal: "Open the Billing page and show the credit balance", values: { query: { value: "pricing", hint: "site search query" } }, verify: async (p) => /billing/.test(await p.url()), }); ``` - `values`: Jev picks which key goes in which field by `hint`; a fill with no matching value escala...