Runs browser tasks and UI tests from a prompt, like jev-ultrafast: Claude writes the task, Jev (TypeSafe) picks every step, Playwright executes it, and a live inspector shows and pauses each decision. Tests are graded by code, never by Jev's DONE; a passing test can be saved as a flow that the stop gate replays with no model when Claude edits UI.
Plugin q-pilot from marketplace q-harness in phuquocchamp/q-harness. Commands read from the repo's manifest — 3 days ago.
flowsHow the q-pilot plugin drives a browser. Use when the user asks to do something in a web browser (search a site, fill a form, find flights; call pilot_run) or types /q-pilot task, after building or changing UI and before saying it works (call pilot_test), when the user types /q-pilot test, when touc
A software engineering team as subagents (product manager, product owner, architect, engineer, database, debugger, reviewer, verifier) that hand work to each other through files, with /q-engine:feature and /q-engine:reproduce-bug workflows and coding-guideline, commit and Playwright CLI skills.
5 skills · 8 agents
Context advice at turn boundaries: focuses Claude Code's own autocompact on the plan, decisions and changed files, and suggests compacting, clearing for a new task or switching model, with Jev (Haiku fallback) making the judgement calls. Starts in shadow mode.
Shows the subagents, teammates and Workflow runs of a session as a live graph: a `q-graph` button above the prompt, an alert when one waits for you, and a pane with the agent tree, what each agent used (tools, skills, MCP servers, rules) and the session's harness.
Draws each Bash call and each file change (Write, Edit, NotebookEdit) as a bordered panel, and frames the code blocks in replies with copy and collapse buttons.
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.