This file provides concrete examples and starter templates for Codex Workspace.
Use these examples when:
- creating new repo metadata files
- scaffolding initial workspace data
- implementing manifest reading in Workspace Hub
- testing direct, external, and mapped-host preview modes
- evaluating optional spec-driven or preview-tool additions
This file complements the prose handover files by showing actual example structures.
A starter template for per-repo runtime metadata.
Recommended destination inside a repo:
.workspace/project.json
A sample shared workspace inventory file.
Recommended destination:
shared/repo-index.json
The tracked default repo-group manifest for targeted update workflows.
A sample repo-group manifest shape for adapting or replacing the default.
The tracked registry for reviewed external sources that graduate into optional abilities or core workspace services.
Starter material for a tracked spec-driven layer when a change is large enough to justify requirements, design notes, and task breakdowns.
Guidance for repo-local component previews using lightweight story files rather than a workspace-wide preview platform.
Starter material for creating a minimal repo README.md and a placeholder cover image when a repo lands under repos/ without enough onboarding context.
This template set also includes a dedicated site-reference README template for public site clone or rip requests, where the repo should be framed as a mirror, reference copy, or rebuild rather than as the original source project.
Guidance for local task artifacts such as plans, summaries, logs, screenshots, and other outputs under cache/context/agents/jobs/.
Guidance for optional local workflow-state layers such as .cognetivy/, without turning them into the tracked source of truth.
Guidance for optional AGENTS.md fragment composition when a single file grows too large to maintain cleanly.
Starter material for official repo-local Codex setup such as .codex/config.toml and .codex/skills/.
Starter material for optional .opencode/ and mixed-tool agent presets without turning OpenCode or oh-my-openagent into workspace dependencies.
The manifest template demonstrates:
- a direct-run frontend project
- a normal slug
- package manager and commands
- direct preview URL
- optional mapped-host path (
mappedHostPathin manifests) - optional healthcheck URL
- lightweight notes and tags
Not every field is required in every real project.
The repo-docs template demonstrates:
- a minimal repo
README.mdstructure - the Workspace Hub cover block markers
- a repo-local
docs/cover.pngpath that can start as a placeholder and later be replaced by a captured preview - a concise checklist for setup, run, preview, and structure notes
- a separate public-site reference template that records source URL, capture date, fetch method, local serve path, mirror-versus-rebuild status, and repo-local
ref/fallback handling for blocked downloads
The sample repo index demonstrates a mixed workspace containing:
- the Workspace Hub itself
- a WordPress site
- a Three.js/WebGL repo
- a static site
These examples show how different repo types can coexist inside the same workspace while using different preview modes.
The repo-group manifest shows how named sets such as core, frontend, or wordpress can be updated selectively without hard-coding organization-specific arrays into scripts.
The capability manifest shows how reviewed upstream sources can be classified and managed without blurring together references, optional abilities, repo-local adoption, and core workspace services.
Use one batching convention across the workspace:
- workspace-wide larger batches live in
docs/HANDOVER.md - repo-specific larger batches live in tracked
openspec/changes/.../tasks.mdor a repo-localNext Batchessection
That keeps the next end-to-end slice visible without turning small edits into mandatory spec work.
When creating new repos or manifests:
- use the template as a starting point
- remove empty fields if they are not useful
- prefer explicit values where runtime behaviour matters
- keep repo metadata lightweight and readable
Use for:
- Vite
- Three.js
- WebGL
- most static/dev-server projects
Use for:
- Local-managed WordPress sites
- repos opened through another tool or service
Use when:
- a clean mapped path or local-domain route is stable
- proxying adds real convenience
- the project has been tested successfully in proxy mode
Use a tracked spec layer only when the change is large enough that intent should outlive one chat session.
Suggested tracked structure:
openspec/
├── specs/
│ └── <capability>/
│ └── spec.md
└── changes/
└── <change-id>/
├── proposal.md
├── design.md
└── tasks.md
This is the cleanest way to add a tracked spec layer without making it mandatory for every small edit.
Do not require it for every small edit.
For frontend repos, prefer repo-local preview tools over a workspace-wide component platform.
Recommended choice:
Ladlefor fast React or Vite component previews with minimal setupStorybookwhen richer docs, addons, or a published component docs surface matter more than setup weight
Keep stories repo-local and prefer Component Story Format when possible so files stay portable between tools.
For longer-running local work, keep transient evidence in cache instead of tracked docs by default.
Suggested local bundle:
cache/context/agents/jobs/<job-id>/
├── plan.md
├── summary.md
├── sources.json
├── logs/
├── screenshots/
└── outputs/
Use:
tools/templates/artifacts/for the conventiontools/scripts/init-agent-job-bundle.shto create a bundle quickly
If a user wants more operational history than a local cache bundle, an optional workflow-state layer such as .cognetivy/ can hold runs, events, and collections.
Use that as local operational state, not as a replacement for tracked docs, specs, manifests, or skills.
This examples file is complete when Codex can use the included template and sample data as working references for:
- manifest creation
- repo inventory scaffolding
- Workspace Hub parser implementation
- preview-mode testing
- optional spec-driven planning
- local artifact-bundle setup