Wire FAI Context

Connect standalone primitives (agents, skills, instructions) to FROOT knowledge modules, WAF pillars, and solution plays via the FAI Protocol wiring layer.

Manifest vs. Context

Aspect	fai-manifest.json	fai-context.json
Level	Solution play (top-level)	Individual primitive
Location	solution-plays/NN-name/	Next to any agent, instruction, or skill
Purpose	Wire ALL primitives for a play	Declare context for ONE primitive
Analogy	Docker Compose (full stack)	Single Dockerfile (one service)
Required	Yes, for every play	No, optional for standalone primitives

:::tip Rule of Thumb Building a solution play? → Create fai-manifest.json.
Building a standalone primitive that needs context? → Create fai-context.json. :::

Creating fai-context.json

Place the context file in a subfolder named after the primitive:

agents/
  fai-rag-architect.agent.md       # The agent itself
  fai-rag-architect/
    fai-context.json                # Context for this agent

agents/fai-rag-architect/fai-context.json

{
  "assumes": [
    "R2-RAG-Architecture",
    "O3-MCP-Tools-Functions",
    "T2-Responsible-AI"
  ],
  "waf": [
    "security",
    "reliability",
    "cost-optimization"
  ],
  "compatiblePlays": [
    "01-enterprise-rag",
    "21-agentic-rag"
  ],
  "evaluation": {
    "groundedness": 0.95,
    "coherence": 0.90
  }
}

The assumes Field

Declares which FROOT knowledge modules your primitive needs:

Series	ID	Module Name
Foundations	F1	GenAI Foundations
	F2	LLM Selection
	F3	AI Glossary
Reasoning	R1	Prompt Patterns
	R2	RAG Architecture
	R3	Deterministic AI
Orchestration	O1	Semantic Kernel
	O2	Agent Coding
	O3	MCP Tools
	O5	GPU Infrastructure
Transformation	T1	Fine-Tuning + MLOps
	T2	Responsible AI
	T3	Production Deploy

Creating fai-manifest.json

Top-level wiring for solution plays:

solution-plays/01-enterprise-rag/fai-manifest.json

{
  "play": "01-enterprise-rag",
  "version": "2.0.0",
  "context": {
    "knowledge": ["R2-RAG-Architecture", "R1-Prompt-Patterns"],
    "waf": ["security", "reliability", "cost-optimization"],
    "scope": "Enterprise RAG with Azure AI Search and GPT-4o"
  },
  "primitives": {
    "agents": ["../../agents/fai-rag-architect.agent.md"],
    "instructions": ["../../instructions/python-waf.instructions.md"],
    "skills": ["../../skills/fai-build-genai-rag/"],
    "hooks": ["../../hooks/fai-secrets-scanner/"],
    "guardrails": {
      "groundedness": 0.95,
      "coherence": 0.90,
      "safety": 0.99
    }
  },
  "infrastructure": {
    "bicep": "./infra/main.bicep"
  }
}

Context Resolution Chain

1. Read fai-manifest.json
   ├── Load context.knowledge[] → FROOT modules
   ├── Load context.waf[] → WAF pillar rules
2. Resolve each primitive reference
   ├── Check for fai-context.json → merge knowledge + waf
   └── Verify compatiblePlays includes this play
3. Apply guardrails from manifest
4. Report: all primitives wired, modules loaded, pillars enforced

Conflict resolution: Knowledge modules and WAF pillars are unioned. Guardrails — manifest wins.

Validation

# Validate all primitives
npm run validate:primitives

# Load play in FAI Engine
node engine/index.js solution-plays/01-enterprise-rag/fai-manifest.json --status

Troubleshooting

Problem	Fix
"0 modules loaded"	Add FROOT module IDs to context.knowledge
"Primitive not found"	Paths are relative to manifest location
Context file ignored	Must be in subfolder named after the primitive
JSON parse error	Validate: node -e "require('./path/file.json')"

See Also

• FAI Protocol — full protocol specification
• Primitives Overview — all 6 primitive types
• Create an Agent — agents with context