Instructions file CodexOpenCode
Repository-specific instructions for tentenco/shuorenhua-zh-tw, including Traditional Chinese writing conventions, project layout, and collaboration rules. The repository is a code project managed with Git.
Instructions file
A writing tool for Traditional Chinese used in Taiwan. It removes AI-like wording, replaces unsuitable regional terms, fixes simplified-to-traditional conversion errors, and applies Taiwan-style punctuation and publishing language.
Skill Claude CodeCodex
A writing and editing guide for making Chinese and English text sound natural and less machine-generated, with special attention to Taiwan Traditional Chinese usage. It preserves technical terms, facts, and the responsible subject.
Skill Claude CodeCodex
Write editorial and marketing copy in the house voice of Aman Resorts — restrained, sensory, place-first luxury travel writing. Use when the user asks for content in "Aman style", "Aman voice", "Aman tone", or wants quiet-luxury / understated-premium editorial: hotel and resort features, destination essays, wellness…
Skill Claude CodeCodex
Automate short-form video editing by having an AI coding agent (Codex, Claude Code, or similar) clone a reference viral video's shot structure onto your own footage — auto-detecting shot cuts with FFmpeg, visually matching clips from your asset library, generating TTS voiceover + subtitles, and outputting an editable…
Skill Claude CodeCodex
A writing workflow for producing long Traditional Chinese articles about design, architecture, hotels, products, and people in the style of Kura Yang and MOT TIMES, a Taiwanese design and culture publication.
Skill Claude CodeCodex
Write a Tenten-style blog article from a topic brief, raw draft, pasted text, or a social/article URL (X/Twitter, Threads, LinkedIn, Medium, Substack, Facebook, Instagram, YouTube). Produces a zh-TW long-form article plus a natively written English version, with Taiwan publishing conventions, AI-tell removal…
Skill Claude CodeCodex
Collect the same question from multiple AI product web interfaces, evaluate the answers with a local frontier judge and adversarial challenge, synthesize the best supported answer, and archive separate standalone HTML scorecard and best-answer reports. Use for AI answer comparisons, model bake-offs, research panels…
Skill Claude CodeCodex
Use ChatGPT Pro Web as an external senior engineer while Codex remains the repository owner, integrator, and independent QA. Use this skill whenever the user asks Codex to operate ChatGPT Pro, use GPT Pro through the web UI for coding, build a two-agent workflow, connect a local repository to ChatGPT Developer Mode or…
Skill Claude CodeCodex
Design, pilot, run, and improve bounded Claude Code agent harnesses using the right primitive for the job: /loop for recurring checks, /goal for completion conditions, hooks for deterministic lifecycle enforcement, dynamic workflows for large finite fan-out, and worktrees for write isolation. Use when a user asks to…
Skill Claude CodeCodex
Convert a piece of content — an article, notes, a prompt, a workflow/SOP, a tutorial, a debugging session, an expert explanation, a failure case, or a conversation trace — into a reusable, scoped, testable Agent Skill. Use this WHENEVER the user says "turn this into a skill", "make this repeatable", "codify this…
Skill Claude CodeCodex
Bootstrap a Claude Code project with the "Fable 5 orchestrator + Codex executor" architecture — verify/install the OpenAI Codex plugin, set Fable 5 as the main model, create deep-reasoner (Opus), fast-worker (Sonnet), and codex-executor sub-agents, and write the orchestration workflow into CLAUDE.md. Use when the user…
Agent
Delegate implementation work to the OpenAI Codex CLI to save Claude tokens. Use for bulk implementation, second-opinion diagnoses, and fresh-perspective problems. Treat Codex as a peer senior engineer, not a reviewer. Runs codex exec non-interactively and returns only the outcome.
Agent
Use for reasoning-heavy phases — architecture decisions, debugging complex issues, algorithm design, tricky trade-offs. Think thoroughly, return a concise conclusion the orchestrator can act on.
Agent
Use for mechanical tasks — boilerplate, tests, formatting, renames, simple edits, repetitive changes across files. Execute efficiently.
Skill Claude CodeCodex
Run complex development tasks as a multi-model orchestration: the main model (Fable 5, max reasoning) acts as tech-lead orchestrator and delegates to three executors — deep-reasoner (Opus) for reasoning-heavy phases, fast-worker (Sonnet) for mechanical work, and Codex CLI as a peer senior engineer with a different…
Skill Claude CodeCodex
Use before designing or implementing a new app, feature, integration, library, infrastructure component, agent workflow, or major subsystem. Forces internal reuse checks, Agent Skill discovery, GitHub/package research, candidate evaluation, and a USE/ADAPT/HARVEST/BUILD decision before coding. Skip only for trivial…
Skill Claude CodeCodex
Turn a rough idea, a requirements list, or a painful existing prompt into ONE copy-paste-ready "Enhanced Prompt" for the token-saving multi-model orchestration workflow in Claude Code (Fable 5 orchestrator + deep-reasoner/Opus + fast-worker/Sonnet + Codex peer). Asks 2-3 clarifying questions in Traditional Chinese…
Skill Claude CodeCodex
Turn a messy spoken product or software idea into a confirmed, testable specification and a clean handoff for a coding agent. Use when the user wants to think aloud, brain-dump, reconstruct intent, create a PRD, or prepare implementation work; do not add this discovery flow to a simple request that is already clear.
Skill Claude CodeCodex
Turn a single company URL into an exhaustive competitor landscape report — a value-ranked comparison table (30–50 companies) plus a Branding / Marketing / Sales strategy read. TRIGGER whenever the user pastes or names a company website and wants competitors, alternatives, rivals, a competitive landscape, market map…
Skill Claude CodeCodex
Run a complete Go-To-Market audit that turns a raw product or monetization idea into a battle-tested GTM plan - evidence reconnaissance, a 12-station decision-tree grilling (strategic role, ICP, beachhead market, delivery/trust, pricing, build sequencing, channels, hidden landmines, brand+domain, kill criteria, ops…
Skill Claude CodeCodex
Stand up a role-based "AI operations team" for social-media / SNS content so the AI does the operating work (strategy, research, planning, copywriting, traffic + sales design, task tracking) and the human only reviews and approves. Use this WHENEVER the user wants to "build an SNS/social-media team", "组建运营 团队"…
Skill Claude CodeCodex
Automated internal link building for Tenten content via the Tenten Content Index MCP (index.tenten.dev). Use whenever writing or updating any Tenten article — tenten.co blog/learning/shopify/d2c posts, tentenai.com or geo.tenten.co pages — to insert internal links by querying the live database of 2,800+ published…
Skill Claude CodeCodex
Higgsfield image prompt director for Banana Pro, Soul Cinema, and GPT-2. Modes: (0) face lock for new characters — Banana Pro single-pass (default), GPT-2 single-pass (higher fidelity, higher credits), or Soul Cinema two-pass, all on mid-gray seamless (locked default; white on request) with black camisole/tank…