ceos-start-here
Skill0 starsChanged 12 days agoMITView on GitHub
Progressive entry point for the Creative Engineering OS. Explains the toolkit, maps website tasks to a specialist through the deterministic local router, and scales the CEOS workflow to FAST, STANDARD, or FLAGSHIP work. Use first when starting or resuming website work with this toolkit, or when unsure which skill applies.
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Read this page once, then load only the skills needed for the current task and stage.
- Distinctive: line 27
Run the local router for one primary recommendation:
- Distinctive: line 73
Do not also install manual copies of the same skills, or they will load twice.
- Very specific: line 78
Read START-HERE.md in full.
- Distinctive: line 93
No invented proof: no fake clients, numbers, testimonials or awards.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Start here — Creative Engineering OS
You are working with a toolkit for building original, premium websites. It is a method plus a library, not a theme. Read this page once, then load only the skills needed for the current task and stage.
1. What is where
| Path | What it is | Load it |
skills/ceos-* (14) | The studio method — how this team works, stage by stage | By name when the stage starts |
library/skills/* (386) | Third-party reference skills (permissive licences, upstream licence kept) | Only the one you need: read library/skills/<name>/SKILL.md |
library/CATALOG.md | Every library skill by discipline with one-line "use it for" | Skim the section you need |
templates/ | Four tested starters: editorial, motion-heavy, immersive-3d, experimental-graphics | scripts/new-project.sh <template> <dest> |
tooling/ | visual-audit/capture.mjs (screens + axe + console), perf/frame-probe.mjs | During review |
standards/STANDARD.md | The quality floor, scaled to the task | Read when the project starts |
agents/creative-auditor.md | Read-only independent critic | Use for a substantial independent critique |
Paths are relative to the toolkit root (the folder that contains this skills/ directory).
2. Route and scale the task
Run the local router for one primary recommendation:
python3 <toolkit>/scripts/ceos-route.py "Create a wireframe for the signup flow" --project .The 19 routes and their preferred capabilities live in integrations/routes.json; the full matrix and detection rules
are in integrations/ROUTING.md. Add --need-secondary only when the route's explicitly described second contribution
is useful. Use repeated --available <capability-id> flags for host-provided skills that have no discoverable skill
directory. Framework detection uses package metadata and config filenames. React Bits, shadcn/ui, and pmndrs/examples
require an existing React runtime; a framework override cannot bypass the package-metadata check.
The router recommends; it does not load skills, dispatch agents, install packages, or alter provider settings. Load only the selected skill for the current stage. Prefer native platform features and existing project conventions when they satisfy the approved direction; do not add dependencies to satisfy a route. If guidance conflicts with approved project direction or a core CEOS skill, the approved direction wins, then the core skill.
| Mode | Use it for | Workflow |
| FAST | One local change, such as a CSS spacing fix | Check the affected state and run relevant checks. Skip concept options, signature prototypes, and broad QA unless the scope grows. Keep applicable safety, accessibility, keyboard/focus, reduced-motion, and security checks. |
| STANDARD | Normal features and page work; default when scope is unclear | Use relevant CEOS stages, build, inspect changed behavior in a real browser, and run relevant quality checks. |
| FLAGSHIP | Launch, rebrand, immersive site, or other high-impact work | Run the full research, direction, prototype, build, browser review, independent critique, technical review, refinement, and release verification path. |
Escalate FAST when behavior, dependencies, or direction choices appear. Escalate STANDARD for launch/rebrand work, shared
multi-route systems, immersive signature moments, or multi-team handoffs. A --mode override is available.
3. CEOS workflow and engineering ownership
ceos-studio-workflow owns creative stage selection and handoffs. For FLAGSHIP work, establish the concept and owner
decision before bulk sections, and render the signature moment at desktop and phone widths. For STANDARD work, use those
gates when the task makes a direction choice or includes a signature interaction. FAST work does not inherit those gates
automatically.
CEOS owns creative orchestration and direction. Existing project engineering orchestration policy wins; in gstack-led work, gstack owns engineering orchestration and Superpowers is limited to narrow TDD and verification roles. Do not run two orchestrators.
Use it with Claude Code
Install as a plugin, then start from the router:
/plugin install creative-engineering-os --marketplace ohnoshivm/creative-engineering-os
/creative-engineering-os:ceos-start-hereDo not also install manual copies of the same skills, or they will load twice.
Use it with Codex, OpenCode, Antigravity, Cursor, or another LLM
- Clone or download the repository.
- Read
START-HERE.mdin full. - Open this skill, then load only the skill needed when its stage starts.
- For a specialist technique, search
library/CATALOG.mdand open that reference skill.
The installer can link skills into supported clients: python3 scripts/ceos.py install --client all --mcp (plan only,
then add --apply).
4. Rules that make the output world-class (short form of standards/STANDARD.md)
- One idea that belongs to this client, not the category. References are calibration, never templates.
- Type is the main material: few sizes, big contrast, measured line length, deliberate line breaks.
- One signature motion; everything else quiet; designed reduced-motion state.
- HTML first: all content and actions work without JS and without WebGL.
- Evidence or it didn't happen: use rendered screenshots, interaction checks, axe/keyboard, console, and measurements in proportion to the work; substantial page reviews use the project's documented responsive sizes.
- No invented proof: no fake clients, numbers, testimonials or awards.
5. First commands in a fresh checkout
python3 scripts/ceos.py doctor # what's installed / missing
python3 scripts/ceos-route.py "Fix a small CSS spacing issue" --project .
./scripts/new-project.sh editorial ~/Projects/my-site && cd ~/Projects/my-site && npm ci && npm run dev
node <toolkit>/tooling/visual-audit/capture.mjs --url http://localhost:4321 --out docs/QA/first-lookIn this repository16 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- ceos-3d-asset-pipelineSkill
- ceos-art-directionSkill
- ceos-creative-codingSkill
- ceos-creative-directionSkill
- ceos-frontend-architectureSkill
- ceos-interaction-architectureSkill
- ceos-motion-directionSkill
- ceos-performance-engineeringSkill
- ceos-production-qualitySkill
- ceos-realtime-3dSkill
- ceos-shader-engineeringSkill
- ceos-studio-workflowSkill
- ceos-visual-auditSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

