Eunho-J

Fast autonomous workflow — Plan + Execute only (no Debate planning, no Code Review). Aliases ulw/ultrawork.

Unified autonomous workflow from spec to validated code (Debate-First, all 5 phases)

Resource-efficient operation mode

Parallel agent execution with Claude Code's native Agent Teams

Browser automation and E2E testing

Frontend UI/UX development expert skill

Git expert skill. Commits, rebases, history search

Turn off codex-troller workflow mode and return to default Codex behavior

Check how user work requests should be handled first, then route to codex-troller workflow only when mode is on

Start codex-troller with interview-first workflow automatically (troller-on)

Time-boxed or count-boxed forced feedback loop skill for any structured task — research, project work, document creation, code development, or analysis. Enforces continuous self-feedback cycles that synthesize existing information into new hypotheses (convergent) and explore adjacent unknowns (divergent). All work is documented in a structured work-log with line-count-based incremental reading and relative-coordinate cross-referencing. The agent never stops voluntarily before the allocated budget expires. Use when the user requests sustained, iterative deep work with a time or loop-count budget — e.g. "spend 30 minutes researching X", "run 10 feedback loops on this code", "iterate on this design until time runs out".

Produce a one-shot, author-side mock ICML review with context-isolated divergence, evidence-gated convergence, and MCP-tracked completion. Use when reviewing a four-page machine-learning paper package for ICML-style soundness, significance, originality, presentation, missing evidence, related work, figures, or tables; use also when resuming an incomplete Ralph Reviewer run.