Skip to content

Context Engineering with SCS

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 tooling) for creating, validating, and versioning structured context as git-native artifacts.

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

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

Individual YAML files containing structured context for AI systems. Think of them as “config files for AI behavior.”

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
  1. Meta Bundle: SCS specification language itself
  2. Standards Bundles: External compliance (HIPAA, SOC2, CHAI)
  3. Domain Bundle: Company knowledge aggregator (1 per company)
  4. Concern Bundles: Functional areas (Architecture, Security, Clinical)
  5. Project Bundles: Individual initiatives (Prior Auth App, Patient Portal)

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