Oğuzcan
Community@ogiboy · Andromeda
<web developer/>
Agent Skills by Oğuzcan
Showing 29 vetted skills indexed across 1 GitHub repositories.
ultragoal
Creates and executes durable multi-goal plans over Codex goal mode artifacts.
cancel
Detects and cancels active OMX modes with session-aware state cleanup.
design
Creates and maintains a repo-local DESIGN.md as the source of truth for UI/UX decisions.
pipeline
Orchestrates configurable multi-stage pipelines with state persistence and resume support.
autoresearch
Runs a validator-gated research loop with persisted state until a completion artifact passes.
skill
Manages local Codex skills through CLI commands for listing, creating, editing, and syncing.
ask
Queries local Claude or Gemini advisor CLIs and saves reusable markdown artifacts.
deep-interview
Runs Socratic interview rounds with weighted ambiguity scoring to produce execution-ready requirement specs.
plan
Generates structured work plans through interview, direct, consensus, and review modes.
ultraqa
Runs adversarial end-to-end QA cycles with hostile scenario matrices, temporary harnesses, and structured evidence reports.
ultrawork
Orchestrates parallel execution lanes with acceptance criteria and evidence-based verification.
prometheus-strict
Generates interview-driven execution plans through clarification, critique, and synthesis roles.
doctor
Diagnose and fix oh-my-codex installation issues across Codex CLI configuration files.
code-review
Reviews code changes through parallel code-reviewer and architect agent lanes with severity-rated findings.
omx-setup
Installs and configures oh-my-codex across user or project scope directories.
ralph
Loops task execution with parallel agent delegation until architect-verified completion.
wiki
Maintains a persistent markdown project wiki with keyword search and lifecycle capture.
visual-ralph
Implements frontend UI from approved visual references using measured verdict and pixel-diff iteration.
best-practice-research
Gathers official and upstream evidence to produce cited best-practice recommendations.
autoresearch-goal
Coordinates professor-critic research missions with Codex goal-mode state reconciliation.
ai-slop-cleaner
Removes AI-generated code slop through regression-tests-first, smell-by-smell cleanup passes.
team
Orchestrates parallel Codex and Claude CLI workers in tmux panes with shared task state.
ralplan
Runs consensus planning with Planner, Architect, and Critic agents until approval.
analyze
Performs read-only repository analysis and returns ranked explanations with evidence and confidence levels.
Frequently Asked Questions About Oğuzcan
FAQPage SchemaWhat tasks can I automate using Oğuzcan's OMX skills?▼
OMX skills automate strategic planning, Socratic deep interviews, durable multi-goal execution over Codex goal mode, parallel task completion, adversarial e2e QA, code review, anti-slop refactoring, performance optimization, and validator-gated research loops, all producing persistent repo-native artifacts.
Who are the OMX skills designed for?▼
OMX targets web developers and engineers using Codex CLI who need structured orchestration: solo developers wanting autonomous loops via autopilot, and teams needing tmux-based multi-agent coordination, shared task lists, and repo-local DESIGN.md and wiki documentation.
How do I install and run OMX skills?▼
Run the omx-setup skill to configure oh-my-codex using current CLI behavior, then use the doctor skill to diagnose and fix installation issues. Individual skills are invoked by name with argument hints, and the skill manager lists, adds, removes, and searches local skills.
What prerequisites and dependencies do OMX skills require?▼
OMX requires the Codex CLI with goal mode support, tmux for team and worker orchestration, and optionally external advisor CLIs such as Claude or Gemini for the ask skill. Notifications integrate with Discord, Telegram, Slack, or OpenClaw webhooks.
How does OMX handle frontend UI verification?▼
The visual-ralph skill orchestrates frontend implementation from generated references, static references, or live URL targets, looping with built-in visual verdicts and pixel-diff evidence until the output matches, leaving a reproducible design system behind.