Structured Context Specification
What is Structured Context?
Section titled “What is Structured Context?”If you’re building with AI, you know the struggle: agent context ends up scattered across prompts, docs, code, and tribal knowledge. It drifts. It conflicts. It’s impossible to validate or version properly.
Structured Context Specification (SCS) is a community-driven spec (and tools) for creating, validating, and versioning structured context as git-native artifacts.
Think of SCS like version control for AI context — the same way Git manages code changes, SCS manages how AI systems understand your organization’s rules, patterns, and requirements.
The Problem
Section titled “The Problem”AI systems lack governed operating environments. Without explicit, validated context:
- AI improvises based on training data (not organizational rules)
- Behavior varies between sessions and agents
- Compliance is impossible to prove
- Context drifts over time
The Solution
Section titled “The Solution”SCS provides:
- Structured Context Documents (SCDs): YAML/JSON files containing precise, validated context
- Bundles: Versioned, composable packages of SCDs
- Import Resolution: Automatic dependency management for complex context hierarchies
- Git-Native: Version, review, and audit context just like code
Key Concepts
Section titled “Key Concepts”Structured Context Documents (SCDs)
Section titled “Structured Context Documents (SCDs)”Individual YAML files containing structured context for AI systems. Think of them as “config files for AI behavior.”
Bundles
Section titled “Bundles”Collections of SCDs packaged together with metadata, imports, and provenance tracking. Bundles are:
- Versioned: Immutable once published
- Composable: Import other bundles to build complex hierarchies
- Auditable: Full provenance and change history
Bundle Types
Section titled “Bundle Types”- Meta Bundle: SCS specification language itself
- Standards Bundles: External compliance (HIPAA, SOC2, CHAI)
- Domain Bundle: Company knowledge aggregator (1 per company)
- Concern Bundles: Functional areas (Architecture, Security, Clinical)
- Project Bundles: Individual initiatives (Prior Auth App, Patient Portal)
What can SCS enable?
Section titled “What can SCS enable?”- Alignment — Ensure AI systems understand and follow your organization’s standards, patterns, and requirements from the start.
- Compliance — Maintain audit trails and enforce policies across all AI-assisted development activities.
- Reproducibility — Version your context just like your code. Roll back, compare, and evolve your AI collaboration over time.
- Collaboration — Share context across teams, tools, and AI systems. Break down silos and build institutional knowledge.
Why Git-Native?
Section titled “Why Git-Native?”SCS treats context as first-class artifacts that live alongside your code:
- Version control for context changes
- Pull requests for context reviews
- Git blame for context provenance
- Branching strategies for context experiments