dream-loop

Imported from nousresearch/hermes-agent (optional-skills/creative/dream-loop/SKILL.md). Install upstream with npx skills add nousresearch/hermes-agent --skill dream-loop. Copyright stays with the author (MIT).

Dream Loop Skill

An autonomous process for building extremely impressive visuals, especially 3D scenes (games, apps, usually browser three.js/WebGL): generate photorealistic concept art of the ideal result, build it, screenshot the live build, have a judge score screenshot vs concept against a gated ladder, and iterate until convergence. Goal: the most visually stunning result at an acceptable frame rate for the target platform (e.g. 60 fps browser, 120 fps modern mobile).

This skill does NOT cover general web-app functionality, 2D UI design, or non-visual quality — only the visual-fidelity loop.

When to Use

• User says "dream loop" or asks for a game/scene/app built to a very high level of graphical fidelity.
• Follow-up refinement passes on an existing visual product (see Follow-up loops below).

Prerequisites

• Concept art: the image_generate tool. If unavailable, stop and ask the user for a concept image (or an image-generation API to connect to).
• Screenshots: browser_exec — serve the build locally (python3 -m http.server for static builds), then new_tab(url), wait_for_load(), capture_screenshot().
• Judging: vision_analyze (see Judge section for the one-image-per-call workaround).
• Optional: Blender for asset modeling — see the blender-3d-automation skill. delegate_task for parallel asset work and fresh-context judging.

If you don't have the tools needed for the full loop, flag that to the user early and stop.

Quick Reference

Procedure

1. The target concept

If the user provided a description of a game, scene, or app, proceed — don't ask for clarification unless it's too vague to generate concept art from. If they didn't, ask for it.

Put working context/files in .dream-loop/ and gitignore it (unless told otherwise).

2. Concept art

The concept is a realistic, high-quality, impressive target: the look of a current AAA game running in real time. Physically plausible materials (wet stone, brushed metal, cloth, glass) with real roughness and normal detail, correct proportions, atmosphere (fog, haze, rain, dust, volumetric light), cinematic lighting with a clear key and rich shadows. It should NOT be stylized or an artistic rendition — it should look like a true screenshot of the ideal result.

Avoid these failure modes when generating with image_generate:

• Overbaked: photographic clutter, film grain, hundreds of unique small objects, excessive detail on every surface that reads as noise. A real-time build with modeled assets won't match this, and it won't even look good.
• Oversimplified: cartoon or toy look, flat shading, blobby primitive shapes, empty surfaces. Boring; will not impress the user.

Aim for the middle ground: beautiful surfaces and materials that shaders render well, strong atmosphere and lighting, an interesting palette, and focused hero elements with fine detail that draws the eye (not every element fighting for attention).

Prompt for "in-engine screenshot" more than "concept art" and discourage the noisy/grainy look. Review the image with vision_analyze; if it hits a failure mode, pass it back to image_generate in edit mode and ask it to fix the issue. Save it as .dream-loop/concept.png.

If you generated the art (the user didn't supply it), pause and confirm it matches the user's vision before starting the build loop.

3. Time budget

If the user gives a time budget, record the start time and check the clock between rounds. Don't degrade visual fidelity to hit the budget — strive for the absolute best result, and don't rush work to the judge. Parallelize or distribute work (e.g. delegate_task for independent assets) to hit the time goal, but no shortcuts: it's better to hit the time limit with meaningful, beautiful progress than with something broadly complete but ugly.

If no time budget is given, run until an exit criterion — but warn upfront that this may consume a lot of tokens.

4. Build loop

Look at the concept art and implement it in one go, making that first pass count across every tier of the score ladder: composition, textures, lighting, details. Sculpt and model assets carefully (or use external ones if allowed); don't settle for basic procedural elements and flat surfaces unless the art style calls for it. Write intermediate files/plans to .dream-loop/.

• If Blender is installed and the concept involves 3D assets, prefer modeling in Blender (see the blender-3d-automation skill). For complex assets, delegate to subagents via delegate_task.
• If the user allows external assets, prefer them over modeling unless the asset is simple. If unspecified, assume NO external assets from the web.
• Do not be lazy with key environmental details (scenery, flooring, buildings): simple shapes look blocky, shiny, flat, and fake. Tiny details and texturing matter and need custom sculpting.
• Use image_generate for textures, normal maps, skyboxes, etc. — better looking and faster than procedural ones.

5. Screenshot

Serve the build (e.g. python3 -m http.server in the build dir), then via browser_exec: new_tab('http://localhost:8000'), wait_for_load(), allow the scene to settle, capture_screenshot(). Target the same resolution and aspect ratio as the concept art so the comparison is fair. Save as .dream-loop/round-N.png.

6. Self-check before judging

Each time, review the candidate screenshot yourself before submitting. Do not submit half-baked work. Compare screenshot and concept side by side and log an honest assessment of judge-readiness; only submit if confident you've significantly improved the score. Be rigorous and audit every pixel: big stuff (missing/incorrect objects, wrong scale, perspective, positioning) and small stuff (rendering glitches, flat untextured surfaces, ugly lighting, poor contrast, washed-out or oversaturated color, speckles, ugly shadows). Scan surface by surface, object by object, and list findings.

7. Judge

Judging should ideally be done by a fresh subagent with a clean context each round (delegate_task), to keep it objective and cheap. Give the judge the latest screenshot, the concept, and (from round 2 on) the previous round's screenshot and verdict.

Mechanics: vision_analyze takes one image per call. Either have the judge make sequential calls (concept, then screenshot, then compare from memory of its own descriptions), or — better — stitch a labeled side-by-side composite with ImageMagick (convert concept.png shot.png +append compare.png) or PIL and analyze that single image.

Judge prompt:

You are an art director reviewing a real-time render against its concept art. Compare the screenshot to the concept and score it 0-10 using this ladder. The ladder is gated: a frame cannot score above a tier's cap until every requirement of the tiers below it is fully met. Be strict about the gates.

• Tier 1, shape (0-3): camera, framing, composition, and the position and rough scale of every major object match the concept. Layout, not finish: every major element present, in the right region of the frame (within ~10% of frame width/height), at roughly the right size (within ~25%). Right place and vaguely correct outline passes even if edges and surface are wrong; save precision nitpicks for Tier 4. Cap 3 until true.
• Tier 2, light and color (3-5): key light direction and color, overall exposure (no clipping to black or white), shadow depth, palette, contrast, atmosphere. Attend to reflections, glows, etc. Ensure the scene is not too bright or dark relative to the concept. Judge the whole frame, not tiny details (Tier 4). Cap 5 until lighting/reflections/color/ contrast are generally right.
• Tier 3, materials and surfaces (5-7): every surface reads as the right material at a glance: textures, roughness, translucency, wetness, reflections. Assets must not look procedural, blocky, smooth/plastic; frame-dominating elements should be properly sculpted and detailed. Cap 7 until true.
• Tier 4, fine detail (7-9): the small things. Nitpick relentlessly; inspect every little object up close. Layout aligns near-perfectly; materials extremely convincing. Cap 9 until right.
• Tier 5, indistinguishable (9-10): holds up side by side and zoomed in. Nitpick every pixel.

If a previous verdict and screenshot are provided: you are one reviewer in a sequence, not the first. Maintain consistency. First mark each previous directive LANDED, PARTIAL, or NOT DONE against the new screenshot; carry forward anything PARTIAL or NOT DONE. Don't reverse a prior directive unless the result is clearly worse — and if you do, say so and why.

Output format:

1. Score on the first line; "Tier N" (highest fully-passed gate) on the second. 1b. If given a previous verdict: the LANDED / PARTIAL / NOT DONE list.
2. "Blocking:" the specific failures of the next tier's gate. The builder must clear these before anything else counts. Name the element and the change, with magnitudes: "Rocks: replace the stacked ovoid boulders with one continuous fractured slab; cracks 2-5cm wide, dark interiors, add surface texture so they don't look flat/plastic" — not "the rocks look artificial".
3. Then at most 4 further directives from higher tiers, same style, ordered by points recoverable.

No non-actionable feedback ("this looks synthetic") — name the specific causes. Every directive must be actionable this round. Don't round up: if a gate isn't fully passed, the cap holds.

8. Exit criteria

• Score >= 8 and target FPS acceptable: done. Show the user the latest screenshot; ask if they want more iterations.
• Score >= 8 but FPS unacceptable: optimize — lossless wins first, then minimal-visual-impact ones. Re-judge afterwards to confirm no regression.
• Stall approaching (best score hasn't improved a full point in 2 rounds, or the judge named the same gap 3 times): stop incremental tweaks. Step back and ask what about the approach is capping the score. Make one big structural change in a round: swap asset strategy (sculpt in Blender, pull real models/textures/HDRIs if allowed), rewrite the lighting model, rebuild the composition, change the camera. Self-check carefully — big changes break things. Only repeat parameter tuning if you can articulate why it would work this time.
• Stalled (already tried a big structural change, score flat 3 rounds, judge is nitpicking or demanding intractable things like raytracing on a GPU-less machine): stop, tell the user why you're blocked, give options.
• Otherwise: address all or most heavy-hitting gaps this round, not just the top one — rounds are expensive. Prioritize gaps that move the needle most (lighting, textures, mesh detail). Only revert if the score dropped a full point or more; small dips are judge noise, and reverting a whole round throws out good changes with bad. If one change clearly regressed, undo just that change. Loop.

Follow-up loops

When building on an existing product (or the user re-invokes the skill for refinements), don't create new concept art in a vacuum — it may diverge from what exists. Instead capture a live screenshot of the current product and prompt image_generate to render the best possible version of it (current screenshot → AAA-graphics version of the same shot), then use that as the target. Multiple screens can run parallel judge loops if asked, at higher token cost.

Pitfalls

• Overbaked or oversimplified concept art (see failure modes) — fix the concept before building against it.
• Submitting half-baked screenshots to the judge; the self-check gate exists for a reason.
• Screenshot at a different resolution/aspect than the concept — unfair comparison, noisy verdicts.
• Tunnel-visioning on incremental tweaks when the judge says you're off base.
• Judging both images in one vision_analyze call — it takes one image; composite them first.
• Screenshotting before the scene loads/settles — add a wait after wait_for_load() for asset streaming and animation warm-up.

Verification

• .dream-loop/concept.png exists and passed the failure-mode review (and user confirmation, if generated).
• Each round has a screenshot, a logged self-assessment, and a judge verdict with score + tier + directives.
• Exit only via an explicit exit criterion; final screenshot shown to the user with the final score and FPS measurement.

Adapted from dream-loop by Anshu Chimala (MIT). Upstream license vendored as LICENSE.txt.