AI Engineering

Audits, repairs, and rewrites MiniMax H3 prompts into the official structured format. Use after the creative director routes an existing prompt for diagnosis or for fixing formatting, timeline, camera, dialogue, and other issues.

Transforms text-only video ideas into precise MiniMax H3 timelines with audiovisual instructions. Used when no image, video, or audio reference assets are supplied.

Writes traceable six-section reference prompts for the MiniMax H3 video model, ensuring consistent character, object, and scene reuse across image-to-video generation, editing, and continuation.

Removes AI writing signatures (like -mektedir overuse, punctuation inflation, cliché patterns) from Turkish text and, where appropriate, inserts human-like rhythm variations. Use when you want to make AI-generated Turkish content sound natural.

CTO-level architectural advisor that audits AI-generated code for design soundness, resilience, and state management. Use when making design decisions, reviewing architectures, or ensuring system quality.

Builds reusable, specialized AI skills for designing and generating research paper figures from academic literature, with rigorous candidate comparison and optimization.

Helps AI agents interpret vague or incomplete creative briefs by prioritizing taste, feeling, and human experience, ensuring the output fulfills the creator's deeper intent rather than mechanically complying with the literal prompt.

Captures snapshots from a camera so Claude Code can visually verify physical hardware like displays, LEDs, and wiring, catching rendering bugs, polarity issues, and stale states that automated tests miss.

Kimchi turns a raw product idea into build-ready docs by interrogating the user through expert personas and locking scope, APIs, priorities, and handoff notes. Use it to validate, scope, architect, or plan a product before any code is written.

A discipline for directing product launch videos using code, ensuring craft, sync, and production quality. Use for launch films, demos, feature announcements, and sizzle reels.

A CLI wrapper for running local Grok, Kimi Code, and Claude Code CLIs with their strongest default models. Supports concurrent batch jobs, dependent dual pipelines, artifact verification, and retries for AI agent workflows.

Spawn a panel of expert agents to debate any question using a structured protocol with convergence diagnostics. Use for high-stakes decisions requiring multi-perspective analysis.