Loading...
Loading...
Found 34 Skills
OpenAI Codex (CLI / IDE / cloud) の公式リファレンス。 codex CLI, codex exec, AGENTS.md, rules, subagents, prompting, approvals, sandbox, permission profiles, auto-review, config.toml, profiles, MCP 設定, 環境変数, GitHub Action, Codex SDK, Agents SDK 連携, cloud 委譲, administration, roles, provisioning, analytics API, compliance API, Codex Security 脆弱性スキャン, deep scan, triage, findings, SARIF export, threat model, security hardening, cloud / local / worktree 環境, git worktrees, Record & Replay, GitHub / Linear / Slack 連携。
Unified setup hub: project init, tool setup, 2-agent config, harness-mem, codex CLI, and rule localization. Use when user mentions setup, initialization, new projects, workflow files, CI setup, LSP setup, MCP setup, codex setup, opencode setup, 2-Agent setup, PM coordination, Cursor setup, harness-mem, claude-mem integration, cross-session memory, localize rules, adapt rules. Do NOT load for: implementation work, reviews, build verification, or deployments.
Emulate supported AI code-review GitHub Actions locally and print a terminal-only review from portable skill instructions. Use when running /review-action or checking local PR-review feedback before publishing.
Novel Cover Generation. Automatically analyze the genre and style based on the book title and author name, call GPT-Image-2 to generate a professional web novel cover with title and signature; Codex CLI prioritizes using the built-in ImageGen, no separate API Key required. Trigger methods: /story-cover, /封面, "Help me make a cover", "Generate cover image", "Make a novel cover", "Cover design"
Get an independent, read-only review from OpenAI Codex (via the codex CLI, on whatever model and reasoning effort the user names, else Codex's configured default) of uncommitted changes, a branch, specific files or a plan, then verify every finding against the real code yourself, accept/reject/defer each one, implement only the accepted fixes, and report a verdict table. Use whenever the user asks for a Codex review, a second opinion, a cross-check or sanity check by another model or GPT, says 'ask Codex', 'have Codex look at this', 'get a second pair of eyes', or wants an independent reviewer before merging, shipping or executing a change or plan, even if they don't mention this skill by name.
Sets up a Model Router that sends thinking work to a head-chef model and doing work to a line-cook model, through Claude Code on the owner's Claude plan and Codex on their ChatGPT plan, or both slots on whichever one plan they have, with the models in models.yaml. Also use when the owner says "check for newer models", "update guava-router" or "uninstall guava-router". Use when the owner asks for model routing or the CLIs, or when hub setup hands off to the router.
Generate original images for websites and apps (hero images, backgrounds, textures, spot illustrations, og:images, product shots, transparent cut-outs, edits of an existing image) and show them in the chat. Uses Codex's built-in image generation on the user's ChatGPT plan when the Codex CLI is signed in (no API key, no model download). Falls back to local FLUX.1-schnell on Apple Silicon, or to a hosted API key. Use whenever a project needs real imagery instead of stock photos, gray boxes, or placeholder URLs, or when the user asks Claude to 'make an image' or 'use Codex to generate an image'.
Test a candidate OpenSEO skill end to end by running fresh, isolated Codex sessions against the local backend and scoring the reports they save. Use when editing a product skill (seo-audit, keyword-research, ...) and you need evidence that the new instructions produce better output, not just a cleaner file.
Run Claude Code or Codex CLI on Muapi (api.muapi.ai) using Muapi credits instead of a separate vendor account. Use when the user wants a local coding agent to use Muapi as its model provider, asks which Muapi models work with Claude Code or Codex, or needs the base URL, API key variable, provider block or model name for either agent. Covers Claude Code (Anthropic Messages) and Codex CLI (OpenAI Responses), model discovery, verification and troubleshooting.
GPT (GPT-6 Astra by default) as advisor, adversarial reviewer or judge through headless `codex exec`, always read-only; never an implementer. Modes: advise (independent read on a plan, diff or design, beside Claude's own view, with follow-ups), review (Codex's built-in reviewer on uncommitted work, a branch or a commit), adversarial (tries to break confidence in a change; optional parallel lenses; every finding verified by Claude before it is shown), judge (scores a reviewer's verdict for insistir and goal-loop). Trigger on "second opinion", "segunda opinión", "ask codex/GPT to review", "que codex lo revise", "revisión adversarial", "adversarial review", "que GPT ataque este cambio", "critica este plan", "challenge this design", "judge this review", "usa astra". Falls back to a Claude subagent, labelled same-provider, when Codex is unavailable. Not askcodex (one-shot text or images) and not for writing code.