dream-loop skill
A loop skill: the agent generates a target screenshot, builds toward it, and uses a critic subagent to close the gap until they match
- 1,976 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The agent autonomously writes and runs code and starts a local preview server
- It uses external image generation and can spend many tokens
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/dream-loopDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent autonomously writes and runs code and starts a local preview server
- It uses external image generation and can spend many tokens
This is third-party code. Review the repository files before installing.
What it does
The skill gives the agent a closed loop for building games, apps and especially 3D scenes to a high visual bar. First the agent generates a target frame via image generation, then builds the project aiming at that frame. A separate critic subagent compares the live screenshot to the target and returns feedback; the agent rebuilds until the critic is satisfied, and can raise the bar by regenerating the target. It has Plus and Pro modes for different subscription tiers, Blender work via MCP or scripting, and batch generation through the fal service. It ships batch-generation scripts and a local preview server.
Who it is for. For developers and designers who want a high visual bar in games, demos and apps.
Good fit when
- You need a demo or scene built to a high visual bar
- You want the agent to iterate the visuals toward a target on its own
- You have a target image and need the live result to match it
Not a fit when
- The agent has no image generation, vision or subagents
- You need logic or backend rather than the visual side
Example request
Build a Three.js graphics demo: isometric camera, voxel style with realistic lighting, a living scene, one hour time limitLimitations
It needs an agent with image generation, built-in or via API such as a Gemini key, with image input and preferably subagents. It was mainly verified with GPT-6 Astra in Codex; other strong models are said to likely work but are not guaranteed. Custom 3D modeling needs Blender. The loop can run long and spend many tokens, and quality depends on the connected generation model.
How to disable. Remove the dream-loop folder from the agent's skills directory. The loop's working files live in the project's .dream-loop folder, which you can delete separately.
FAQ
What does it need to work?
An agent with image generation, image input and preferably subagents. Blender helps for custom 3D modeling.
Why does the loop spend many tokens?
The agent rebuilds until the critic is happy and may regenerate the target. Without a time limit the README warns of heavy token use.
Discussion
No questions or reviews yet.
Related
SkillsA skill that turns your narration and real assets into a 20–60 second video in the style you pick, rendered locally as MP4.
A skill that writes prompts for video, image and music generation on 14 platforms and can submit them in the browser
A director skill: it interviews you about a video idea and turns it into a second-by-second shot script for HyperFrames
Skills and a CLI for video editing, motion graphics, 3D scenes and design, where the agent works in one local project.
Two skills for generative video and images: video writes Seedance, Kling and Veo prompts by dramaturgy rules, image picks the model and its syntax
A skill that turns a plot or scene into a cinematic AI video prompt for 6 to 30 seconds, with shots, performance and sound.
README in short
The README describes dream-loop as a closed loop: the agent dreams up a target screenshot via image generation, builds the project toward it, a separate critic compares the live frame to the target and gives feedback, and this repeats until they match. Install via npx skills add or by copying into the skills directory. It needs image generation, image input and preferably subagents, plus Blender for 3D modeling. It notes the main testing was with GPT-6 Astra in Codex and gives a live demo example. MIT license.
README badge
Original SKILL.md
--- name: dream-loop description: Build a game or app from a description so that a live screenshot matches a generated rendering. Use when the user says "dream loop" or asks for something built to a very high level of graphical fidelity. license: MIT --- # dream-loop This is a process for you to autonomously build extremely impressive visuals, especially for 3D scenes (e.g. in a game or app). Your first step is to determine which workflow to use: - If the user told you to use the Plus or Pro workflow explicitly, that's your answer - If not, check if the user is on a low or high tier coding subscription. "Low tier" means ChatGPT Plus or equivalent. "High tier" means ChatGPT Pro. For low-tier subscriptions, use the Plus workflow. For high-tier, use the Pro workflow. Plus workflow: read [references/plus-mode/workflow.md](references/plus-mode/workflow.md) Pro workflow: read [references/pro-mode/workflow.md](references/pro-mode/workflow.md) Do not read both documents. They are not inter-compatible.