Agent Skill
Version-matched guidance for turning a creative brief into the right operations.
slopcamera skill install --target agents
Slopcamera
Slopcamera is a local visual studio for coding agents. Author scenes, combine generated and recorded media, and export images, diagrams, animation, and video from retained sources.
Free and open source under the MIT license · Source build · Bun 1.3.14 or newer · Recording needs macOS
$ slopcamera diagram init first.diagram.json
# Create the included editable diagram in a new working directory.
$ slopcamera diagram check first.diagram.json --strict
# Validate its objects, labels, and layout.
$ slopcamera diagram render first.diagram.json
# Write example-flow.tldr and light/dark SVG and PNG files.
Slopcamera currently installs from source. Git and Bun 1.3.14+ are required.
Start a new source checkout
git clone --branch main https://github.com/hraness/slopcamera.git slopcamera-sourcecd slopcamera-sourceAfter building: install the matching Agent Skill
bun apps/desktop/dist/cli/main.js skill install --target agents
For Claude Code: bun apps/desktop/dist/cli/main.js skill install --target claude
Cloning is the first step. Follow the source guide to record the commit, install locked dependencies, build the SDK and CLI, and define the slopcamera command. Then run the skill command above from that checkout.
Read the complete source-install guide and packaged Agent Skill. No Slopcamera release archive has been published.
Name the source, result, and details to preserve. Your agent plans the commands and shows a preview before final delivery.
“Create a product reveal with a moving camera. Keep the model editable, mount the product diagram on a screen in the scene, and show contact frames before rendering.”
Portable Three.js scenes, supported GLB assets, an explicit hardware GPU profile, and saved splat worlds through Spark. Splats capture appearance, not collision geometry. Direct a scene.
“Build a shaded street with an original presenter. Put an animated explanation on a screen in the world, then pull the camera back into the city.”
Blender scenes, character rigs, cloth and liquid caches; CadQuery solids and STEP. Native studio commands require the current source build and separately installed engines. Make a native film.
“Turn the services in this repository into an editable diagram, then animate the explanation. Add a presenter, readable mathematics, and captions where they help.”
Diagrams, HTML/SVG motion, shader graphics, and optional vgpu raster passes. Manim educational films use the source-build studio. Make an educational video.
“Record my product demo, remove pauses, zoom into important clicks, and add
logo.svg. Generate a calm voiceover fromscript.txt, then preview clean and captioned versions in 16:9, 9:16, 1:1, and 4:5.”
Edit screen recordings and imported footage, or generate images, video, and voice through your own Gateway access. Source-build shot recipes retain budgets, reviewed takes, and continuation references. Edit video · Generate media · Direct takes.
Source availability: Build from main for portable scenes, the qualified Three.js GPU profile, saved worlds, native studio, shot-recipe direct, and scene camera-track. Engines install separately. Check capabilities and requirements.
Review the result and its source, preview, and final paths. Ask for changes in plain language.
Keep native rigs and simulations, portable geometry and cameras, diagram objects and labels, and video edits in their respective sources.
Share supported geometry and calibrated cameras, or mount images and video in world space. Raster handoffs preserve pixels, without reconstructing geometry or rigs. Understand scene interchange.
Inspect available tools, then author scenes or diagrams, import assets, or record on macOS.
Set cameras, shots, timing, and outputs. Discover Gateway models for generated images, video, voice, or transcripts.
Combine selected shots, edit timing, and add captions and graphics. Reframe footage and treat color and audio.
Inspect contact frames, motion, sound, and continuity. Preview and final renders use the same timeline and composition.
Export finished videos. Retain sources, project decisions, and receipts for revisions.
Start with the Agent Skill and CLI, integrate through the SDK, or use the smaller MCP toolset.
Version-matched guidance for turning a creative brief into the right operations.
slopcamera skill install --target agents
Human-readable commands and stable JSON for local scripts.
slopcamera workflows list --json
Declarative or imperative media work inside your own Bun code.
import { vectorizeImage } from "@hraness/slopcamera"
A fixed toolset for checking and rendering diagrams, vectorizing images, and generating images within one selected root. It does not expose every CLI command.
slopcamera mcp --root /absolute/path/to/workspace
Keep sources, edits, and outputs in your own storage. Rendered derivatives connect native and portable assets without erasing their differences.
Native Python execution requires separate authorization. Loading a custom Bun workflow evaluates its module, including during check and plan; review the source before invoking it. Both run as your current user, without an operating-system sandbox. Runtime hashes identify observed tools; they do not make arbitrary source hermetic. Read the security policy and privacy guide.
Find a path through the documentation index: learn with a first diagram, run a workflow, look up capabilities, or understand the architecture.
No Slopcamera account or subscription exists. Slopcamera is free and open source under the MIT license. Model-backed generation uses caller-owned Vercel AI Gateway access.
Slopcamera is free. Editing, diagrams, vectorization, and exports run locally. Model usage is billed to your Vercel AI Gateway account.
In the local CLI or SDK environment; never stored, printed, or sent to this website. A linked Vercel project can inject it for one command with vercel env run -- <command>.
The CLI needs Bun 1.3.14+ on macOS, Linux, or Windows; recording needs macOS permissions. GPU, browser, codec, and engine requirements vary; slopcamera doctor and the capability reference identify those requirements.
Normal edit operations leave original recordings and imported media unchanged. Diagram rendering replaces its named derived outputs.
Editing and rendering stay local. Gateway generation and selected cloud analysis upload named media only after acknowledgement. Native Python and custom Bun workflows run with your current-user access, including potential network access.
After the source build, slopcamera skill install installs its guide for Codex. Use --target claude for Claude Code or --target agents for other readers. Add --scope project inside a repository for a project-only install.
No. The website explains and installs the local system. Media work runs through local tools; MCP exposes only its documented subset.
Ben Guo, who builds Hraness projects from Puerto Rico. The maker section has the short version and links.
Slopcamera is built by Ben Guo, a musician and builder, formerly a founder and engineering leader at companies including Venmo and Stripe, now building from Puerto Rico.
Install the skill, name the source files and the details that must stay the same, and ask for the finished result.
Free and open source under the MIT license · Source build · Bun 1.3.14 or newer