Handbook
LCDL human-first information architecture
LCDL documentation must answer a human question before it exposes a repository structure. A reader should know where to go based on intent, not file paths.
Updated
Principle
LCDL documentation must answer a human question before it exposes a repository structure. A reader should know where to go based on intent, not file paths.
Primary reader intents
| Intent | UX entry point | Do not show first |
|---|---|---|
| “What is this?” | Start → What is LCDL | Contract pages, package layout |
| “Can I run it?” | Start → Install / First task | Enterprise operations |
| “Can I learn it?” | Learn → 101 / 201 / 301 | API reference |
| “Can I build with it?” | Build → execution, RAG, repair, operators | Generated contract dumps |
| “Can I govern it?” | Enterprise → architecture, security, operations | Low-level helper list |
| “Can agents use it?” | Agents → Cursor / MCP / workflows | Internal .cursor file tree |
| “What exactly is the schema/API?” | Reference → task catalog, schemas, API | Onboarding narrative |
Canonical nav tree
Home
Start
What is LCDL?
Choose your path
Install
First governed task
Troubleshooting
Learn
101 Foundations
201 Building with LCDL
301 Enterprise operation
Examples
Build
Execution
Operators
RAG and evidence
Verification and repair
Page intelligence
Games
Reference
API
ContractSpec
Schemas
Task catalog
Errors
Operator reference
Enterprise
Architecture
Security and responsible use
Governance and auditability
Operations
Benchmarks
Production readiness
Agents
Cursor
MCP sidecar
MCP client
Agent workflows
Prompt snippets
Contribute
Docs authoring
Task authoring
Release checklist
Website publishing
Rule
The site can generate HTML for every Markdown file, but only curated pages appear in the primary top nav and active-section sidebar. Deep contracts and generated catalogs are reachable from hubs.