VCP Newcomer's Guide

For: Researchers curious about the Value-Context Protocol Date: 2026-01-12 Version: 1.0

What is VCP?

The Value-Context Protocol (VCP) is a unified protocol stack for expressing, transporting, and applying constitutional values to AI systems. Think of it as the "OSI model for AI ethics" — a layered architecture where each layer handles a specific concern.

The core insight: AI alignment requires not just encoding values, but a shared language for exchanging, comparing, and negotiating those values between humans, AI agents, and institutions.

The Six Layers

┌─────────────────────────────────────────────────────────────────┐
│  Layer 6: VCP-ECONOMIC GOVERNANCE (VCP/E)                       │
│  "Who Pays" - Fiduciary constraints, authorization gaps         │
│  Handles: Transaction governance, capability/accountability     │
├─────────────────────────────────────────────────────────────────┤
│  Layer 5: VCP-MESSAGING (VCP/M)                                 │
│  "Who Talks" - Inter-agent message exchange, escalation         │
│  Handles: Context sharing, constitution announce, escalation    │
├─────────────────────────────────────────────────────────────────┤
│  Layer 4: VCP-ADAPTATION (VCP/A)                                │
│  "When and How" - Context encoding, state tracking              │
│  Maps to: VCL (Values Communication Layer)                      │
├─────────────────────────────────────────────────────────────────┤
│  Layer 3: VCP-SEMANTICS (VCP/S)                                 │
│  "What it Means" - CSM1 grammar, personas, composition          │
│  Maps to: CSM (Constitutional Safety Minicode)                  │
├─────────────────────────────────────────────────────────────────┤
│  Layer 2: VCP-TRANSPORT (VCP/T)                                 │
│  "How it's Delivered" - Signed bundles, verification, audit     │
│  Maps to: Secure transport, proof-of-adherence                  │
├─────────────────────────────────────────────────────────────────┤
│  Layer 1: VCP-IDENTITY (VCP/I)                                  │
│  "What it's Called" - Token naming, namespaces, registry        │
│  Maps to: UVC (Universal Values Corpus)                         │
└─────────────────────────────────────────────────────────────────┘

Mapping to Research Proposal

Research Concept	VCP Layer	Purpose
UVC (Universal Values Corpus)	Layer 1: Identity	Curated cross-cultural corpus of values, "What3Words for ethics"
CSM (Constitutional Safety Minicode)	Layer 3: Semantics	Compact grammar for safety rules, adherence proofs
VCL (Values Communication Layer)	Layer 4: Adaptation	Emoji-based symbolic encoding for context
VCP/M (Messaging)	Layer 5: Messaging	Inter-agent message types, escalation, delivery semantics
VCP/E (Economic Governance)	Layer 6: Economic	Fiduciary constraints, authorization, transaction governance

Layer 1: VCP-Identity (VCP/I)

Question answered: "What is this constitution called?"

Token Format

Values are identified using dot-separated tokens:

{domain}.{approach}.{role}@{version}

Examples:
  family.safe.guide@1.2.0
  business.ethical.advisor@2.0.0
  healthcare.caring.assistant@1.0.0

URI Scheme

Full addressable URIs for value references:

creed://{authority}/{token}

Example:
  creed://creed.space/family.safe.guide@1.2.0

Key Files

File	Purpose
`services/vcp/identity/token.py`	Token parsing and validation
`services/vcp/identity/namespace.py`	Namespace management
`services/vcp/identity/registry.py`	Value registry operations
`docs/identity/VCP_IDENTITY_NAMING.md`	Naming specification

Layer 2: VCP-Transport (VCP/T)

Question answered: "How is it delivered securely?"

Purpose

VCP/T ensures constitutions are:

Authentic: Verified origin (not tampered)
Auditable: Every decision traced
Versioned: Changes tracked over time

Signed Bundles

Constitutions are packaged with cryptographic signatures:

{
  "constitution": {
    "id": "family.safe.guide",
    "version": "1.2.0",
    "rules": [ ... ],
    "persona": "NANNY",
    "adherence": 5
  },
  "vcp": {
    "token": "family.safe.guide@1.2.0",
    "csm1": "N5+F+E",
    "uri": "creed://creed.space/family.safe.guide@1.2.0"
  },
  "signature": {
    "algorithm": "ed25519",
    "value": "base64:...",
    "authority": "creed://creed.space",
    "timestamp": "2026-01-12T00:00:00Z"
  }
}

Verification Chain

┌─────────────────────────────────────────────────────────────┐
│  AUTHORING                                                   │
│  Constitution created → Signed by authority                 │
└─────────────────────────┬───────────────────────────────────┘
                          │
                          ▼
┌─────────────────────────────────────────────────────────────┐
│  TRANSPORT                                                   │
│  Bundle transmitted via API/export/MCP                      │
└─────────────────────────┬───────────────────────────────────┘
                          │
                          ▼
┌─────────────────────────────────────────────────────────────┐
│  VERIFICATION                                                │
│  Consumer verifies signature against trusted authorities    │
│  If invalid → REJECT (fail-closed)                          │
└─────────────────────────┬───────────────────────────────────┘
                          │
                          ▼
┌─────────────────────────────────────────────────────────────┐
│  AUDIT                                                       │
│  Decision logged: constitution, context, outcome, timestamp │
│  Retention: GDPR-compliant, configurable TTL                │
└─────────────────────────────────────────────────────────────┘

Trust Anchors

Authority	Trust Level	Use Case
`creed://creed.space`	ROOT	Official Creed Space constitutions
`creed://org.example`	DELEGATED	Organization-specific policies
`creed://user.{id}`	USER	Personal constitutional preferences

Audit Trail

Every PDP decision is logged:

{
  "decision_id": "uuid",
  "timestamp": "2026-01-12T04:30:00Z",
  "constitution_hash": "sha256:...",
  "vcp_context": "⏰🌅|📍🏡|👥👶",
  "outcome": "ALLOW",
  "persona": "NANNY",
  "adherence": 5,
  "plugins_triggered": ["vcp_adaptation", "epistemic_guardian"],
  "latency_ms": 12
}

Proof of Adherence

VCP/T can attest that a response was generated under specific constitutional constraints:

Creed: 2.1.4 sha256:abc123...

Meaning:
- Constitution version 2.1.4 was active
- Hash abc123 identifies exact ruleset
- Verifiable by any party with the constitution

Key Files

File	Purpose
`services/ed25519_constitution_signer.py`	Ed25519 signing
`services/gpg_policy_signer.py`	GPG signing for policies
`services/constitution_enforcer.py`	Signature verification
`services/safety_stack/audit.py`	Decision audit logging
`services/export_formatter.py`	Bundle generation with VCP metadata
`api_routers/export.py`	Export endpoints

Export Formats

Format	Purpose	VCP Included
Claude Code	System prompt for Claude	✅ Full VCP block
API Config	JSON for API consumers	✅ Full VCP block
MCP Bundle	Model Context Protocol	✅ VCP metadata
Raw YAML	Source constitution	Token only

Layer 3: VCP-Semantics (VCP/S)

Question answered: "What do the rules mean?"

CSM1 Grammar (Constitutional Safety Minicode)

Compact encoding for safety configurations:

Format: {Persona}{Adherence}[+{Scope}]*

Examples:
  N5+F+E    = Nanny persona, adherence 5, scopes: Family, Education
  A3+W      = Ambassador persona, adherence 3, scope: Work
  S5        = Sentinel persona, adherence 5, no extra scopes

Personas

Code	Persona	Role
N	Nanny	Child safety specialist
S	Sentinel	Security guardian
G	Godparent	Balanced mentor
A	Ambassador	Professional representative
M	Muse	Creative collaborator
R	Anchor	Stability provider

Adherence Levels

Level	Meaning
1	Minimal constraints
3	Moderate guidance
5	Maximum safety

Key Files

File	Purpose
`services/vcp/semantics/csm1.py`	CSM1 parser and encoder
`services/vcp/semantics/personas.py`	Persona definitions
`services/vcp/semantics/composition.py`	Rule composition
`docs/semantics/VCP_SEMANTICS_CSM1.md`	CSM1 specification

Layer 4: VCP-Adaptation (VCP/A)

Question answered: "When and how does it apply?"

The Enneagram Protocol

Context encoded across 9 dimensions using emoji:

#	Symbol	Dimension	Example Values
1	⏰	TIME	🌅morning, 🌆evening, 🌙night
2	📍	SPACE	🏡home, 🏢office, 🏫school
3	👥	COMPANY	👤alone, 👶children, 👔colleagues
4	🌍	CULTURE	🇺🇸american, 🇯🇵japanese, 🌍global
5	🎭	OCCASION	➖normal, 🎂celebration, 🚨emergency
6	🧠	STATE	😊happy, 😰anxious, 🤔contemplative
7	🌡️	ENVIRONMENT	☀️comfortable, 🥵hot, 🔇quiet
8	🔷	AGENCY	👑leader, 🤝peer, 🔐limited
9	🔶	CONSTRAINTS	○minimal, ⚖️legal, 💸economic

Wire Format

Compact transmission:

⏰🌅|📍🏡|👥👶

Meaning: morning, at home, children present

State Tracking

Context changes are tracked across interactions:

tracker.record(ctx1)  # morning, home
tracker.record(ctx2)  # evening, office
# → Transition detected: MINOR (time, space changed)

Transition Severity

Severity	Trigger
NONE	No dimensions changed
MINOR	1-2 non-critical dimensions
MAJOR	3+ dimensions OR critical dimension (agency, occasion, constraints)
EMERGENCY	🚨 emoji present anywhere

Key Files

File	Purpose
`services/vcp/adaptation/context.py`	Context encoding
`services/vcp/adaptation/state.py`	State tracking
`services/vcp/adaptation/redis_state.py`	Redis persistence
`docs/adaptation/VCP_ADAPTATION.md`	Layer specification
`docs/VCP_CONTEXT_DATA_FLOW.md`	Full dataflow reference

Dataflow: End to End

┌─────────────────────────────────────────────────────────────────────────┐
│                           REQUEST                                        │
│  User message + metadata: {time_of_day: "morning", audience: "children"}│
└─────────────────────────────┬───────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────────────┐
│                      VCP/A: ENCODE CONTEXT                               │
│  ContextEncoder.encode(time="morning", company=["children"])            │
│  Result: VCPContext → "⏰🌅|👥👶"                                         │
└─────────────────────────────┬───────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────────────┐
│                      VCP/A: TRACK STATE                                  │
│  StateTracker.record(context) → Transition(severity=MINOR)              │
│  Redis: vcp:state:{session_id}:history                                  │
└─────────────────────────────┬───────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────────────┐
│                      PDP: POLICY DECISION                                │
│  VCPAdaptationPlugin emits signals:                                     │
│    vcp_context_wire: "⏰🌅|👥👶"                                         │
│    vcp_has_company: true                                                │
│    vcp_company: ["👶"]                                                  │
│                                                                         │
│  Action: {prefer_persona: "nanny", content_filter: "family_safe"}       │
└─────────────────────────────┬───────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────────────┐
│                      VCP/S: APPLY SEMANTICS                              │
│  Constitution: N5+F (Nanny, adherence 5, Family scope)                  │
│  Behavior: Child-safe mode activated                                    │
└─────────────────────────────┬───────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────────────┐
│                      VCP/T: AUDIT & SIGN                                 │
│  Response signed with decision trace                                    │
│  Audit log: decision, constitution, context, timestamp                  │
└─────────────────────────────┬───────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────────────┐
│                          RESPONSE                                        │
│  AI response with family-safe content, nanny persona characteristics    │
└─────────────────────────────────────────────────────────────────────────┘

Storage & Persistence

Redis Keys (VCP Context)

vcp:state:{session_id}:history

Contents:
[
  {"timestamp": "2026-01-12T04:30:00Z", "context": {"time": ["🌅"], "space": ["🏡"]}},
  {"timestamp": "2026-01-12T05:00:00Z", "context": {"time": ["☀️"], "space": ["🏢"]}}
]

TTL: 1 hour

What IS Stored

Emoji-encoded context signals (non-PII)
Session ID as key prefix
Timestamps

What is NOT Stored

User messages
AI responses
Personal data
Constitution content

HTTP API Endpoints

Endpoint	Method	Purpose
`/api/vcp/token/validate`	POST	Validate VCP token format
`/api/vcp/csm1/parse`	POST	Parse CSM1 code
`/api/vcp/context/encode`	POST	Encode context to wire format
`/api/vcp/status`	GET	Check VCP layer status

Example: Token Validation

curl -X POST /api/vcp/token/validate \
  -H "Content-Type: application/json" \
  -d '{"token": "family.safe.guide@1.2.0"}'

Response:
{
  "valid": true,
  "canonical": "family.safe.guide",
  "domain": "family",
  "approach": "safe",
  "role": "guide",
  "version": "1.2.0",
  "uri": "creed://creed.space/family.safe.guide@1.2.0"
}

Example: Context Encoding

curl -X POST /api/vcp/context/encode \
  -H "Content-Type: application/json" \
  -d '{"time": "morning", "space": "home", "company": ["children"]}'

Response:
{
  "wire_format": "⏰🌅|📍🏡|👥👶",
  "json_format": {"time": ["🌅"], "space": ["🏡"], "company": ["👶"]},
  "dimensions_set": ["time", "space", "company"]
}

MCP Integration

VCP is also available via Model Context Protocol:

mcp-cli call vcp/vcp_status '{}'
mcp-cli call vcp/vcp_validate_token '{"token": "family.safe.guide@1.2.0"}'
mcp-cli call vcp/vcp_parse_csm1 '{"code": "N5+F+E"}'
mcp-cli call vcp/vcp_encode_context '{"time": "morning", "space": "home"}'

Conflict Resolution (Constitution Composition)

When multiple constitutions or rules apply, VCP uses a precedence system.

Composition Rules

┌─────────────────────────────────────────────────────────────┐
│  CONSTITUTION STACK (most restrictive wins)                 │
├─────────────────────────────────────────────────────────────┤
│  1. Platform Safety (UEF - Universal Ethical Fallback)      │
│  2. Organization Policies                                   │
│  3. User Preferences                                        │
│  4. Session Context (VCP/A)                                 │
└─────────────────────────────────────────────────────────────┘

Conflict Types

Conflict	Resolution	Example
Persona clash	Higher adherence wins	N5 overrides A3
Scope overlap	Union of scopes	F+E + W = F+E+W
Rule contradiction	More restrictive wins	"Allow" + "Block" = Block
Context mismatch	Current context wins	Office overrides "home default"

CSM1 Composition

# Two constitutions applied
constitution_1 = "N5+F"      # Nanny, adherence 5, Family
constitution_2 = "A3+W+E"    # Ambassador, adherence 3, Work, Education

# Composed result (higher adherence, union of scopes)
composed = "N5+F+W+E"        # Nanny wins, all scopes active

Implementation

# In services/vcp/semantics/composition.py
from services.vcp.semantics import compose_csm1

result = compose_csm1(["N5+F", "A3+W+E"])
# Result: CSM1Result(persona="NANNY", adherence=5, scopes=["F", "W", "E"])

Conflict Detection

VCP can detect when constitutions have irreconcilable conflicts:

from services.vcp.semantics import detect_conflicts

conflicts = detect_conflicts([constitution_1, constitution_2])
# Returns list of ConflictReport objects if issues found

Inter-Agent Messaging (Future)

VCP/A spec defines a protocol for AI agents to exchange context and negotiate values. Status: Specified but not yet implemented.

Concept

┌─────────────────┐          VCP Message          ┌─────────────────┐
│   Agent A       │  ─────────────────────────▶  │   Agent B       │
│                 │                               │                 │
│  Context: 🏡👶  │    {"context": "⏰🌅|👥👶",  │  Context: 🏢👔  │
│  Persona: N5    │     "constitution": "N5+F",  │  Persona: A3    │
│                 │     "request": "handoff"}    │                 │
└─────────────────┘                               └─────────────────┘
                              │
                              ▼
                    ┌─────────────────┐
                    │  NEGOTIATION    │
                    │  Agent B adopts │
                    │  N5 for session │
                    └─────────────────┘

Message Format (Draft)

{
  "vcp_version": "2.0",
  "message_type": "context_handoff",
  "sender": {
    "agent_id": "agent-a-uuid",
    "current_context": "⏰🌅|📍🏡|👥👶",
    "active_constitution": "N5+F"
  },
  "payload": {
    "request": "adopt_context",
    "reason": "User moving from home assistant to work assistant",
    "proposed_context": "⏰☀️|📍🏢|👥👔"
  },
  "signature": "ed25519:..."
}

Use Cases (Planned)

Scenario	VCP Message
Agent handoff	Transfer context when user switches assistants
Multi-agent task	Coordinate values across collaborating agents
Escalation	Pass context to human reviewer with full state
Audit request	External auditor queries agent's active constitution

Research Questions

How do agents negotiate when constitutions conflict?
What trust model for inter-agent value claims?
Can VCP enable "constitutional diplomacy" between AI systems?

Creed Space Frontend Integration

How users interact with VCP through the Creed Space web interface.

User Journey

┌─────────────────────────────────────────────────────────────────────┐
│  1. BROWSE CONSTITUTIONS                                            │
│     User explores constitution library                              │
│     Each constitution shows VCP token: family.safe.guide@1.2.0     │
└─────────────────────────────┬───────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────────┐
│  2. SELECT PERSONA                                                  │
│     User picks persona (Nanny, Sentinel, Ambassador, etc.)         │
│     Sets adherence level (1-5 slider)                              │
│     UI shows: "Your settings: N5+F"                                │
└─────────────────────────────┬───────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────────┐
│  3. CONFIGURE CONTEXT (optional)                                    │
│     User specifies typical context                                  │
│     "I usually use this at home with my kids"                      │
│     System encodes: ⏰*|📍🏡|👥👶                                    │
└─────────────────────────────┬───────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────────┐
│  4. EXPORT                                                          │
│     User exports constitution for their AI assistant               │
│     Export includes full VCP metadata block                        │
│     Format: Claude Code / API Config / MCP Bundle                  │
└─────────────────────────────────────────────────────────────────────┘

UI Components

Component	VCP Integration
Constitution Card	Shows VCP token, persona icon, CSM1 code
Adherence Slider	1-5 scale, updates CSM1 in real-time
Scope Selector	Checkboxes for F/W/E/H/etc., updates CSM1
Context Builder	Emoji picker for 9 dimensions
Export Modal	Shows VCP block preview before download

Example: Constitution Card

┌─────────────────────────────────────────────────────────────┐
│  👶 Family Safe Guide                                       │
│                                                             │
│  Token: family.safe.guide@1.2.0                            │
│  Persona: NANNY                                            │
│  CSM1: N5+F+E                                              │
│                                                             │
│  "Gentle guidance for family environments with children.   │
│   Prioritizes safety, education, and age-appropriate       │
│   content."                                                │
│                                                             │
│  [Configure]  [Preview]  [Export]                          │
└─────────────────────────────────────────────────────────────┘

Live Preview (Chat Pane)

Users can test their constitution in a chat interface:

┌─────────────────────────────────────────────────────────────┐
│  Chat with: family.safe.guide@1.2.0                        │
│  Context: ⏰🌅|📍🏡|👥👶                                     │
│  Persona: NANNY (N5+F+E)                                   │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  User: Tell me a bedtime story                             │
│                                                             │
│  AI: Once upon a time, in a cozy little cottage...        │
│      [Response filtered through N5 adherence]              │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Export VCP Block

When user exports, they get a complete VCP metadata block:

{
  "vcp": {
    "token": "family.safe.guide@1.2.0",
    "token_canonical": "family.safe.guide",
    "uri": "creed://creed.space/family.safe.guide@1.2.0",
    "csm1": "N5+F+E",
    "persona": "NANNY",
    "persona_description": "Child safety specialist",
    "adherence_level": 5,
    "version": "2.1.0"
  }
}

Key Frontend Files

File	Purpose
`superego-frontend/src/routes/creeds/`	Constitution browsing
`superego-frontend/src/lib/components/PersonaSelector.svelte`	Persona picker
`superego-frontend/src/lib/components/AdherenceSlider.svelte`	Adherence control
`superego-frontend/src/lib/components/ExportModal.svelte`	Export with VCP
`superego-frontend/src/lib/components/ChatPane.svelte`	Live testing

Feature Flags

Flag	Default	Purpose
`vcp_identity_enabled`	ON	VCP/I layer active
`vcp_semantics_enabled`	ON	VCP/S layer active
`vcp_adaptation_enabled`	ON	VCP/A layer active
`vcp_full_stack_enabled`	ON	All layers enabled
`vcp_redis_persistence_enabled`	ON	Cross-worker state via Redis
`vcp_strict_mode`	OFF	Strict validation mode
`vcp_*_shadow`	OFF	Shadow mode (signals only, no enforcement)

Override via environment:

FF_VCP_ADAPTATION_ENABLED=false  # Disable VCP/A

Directory Structure

services/vcp/
├── __init__.py              # Package exports
├── identity/                # Layer 1: VCP/I
│   ├── token.py             # Token parsing
│   ├── namespace.py         # Namespace management
│   ├── registry.py          # Value registry
│   └── encoding.py          # Multiple encodings
├── semantics/               # Layer 3: VCP/S
│   ├── csm1.py              # CSM1 grammar
│   ├── personas.py          # Persona definitions
│   └── composition.py       # Rule composition
├── adaptation/              # Layer 4: VCP/A
│   ├── context.py           # Context encoding
│   ├── state.py             # State tracking
│   └── redis_state.py       # Redis persistence
└── transport/               # Layer 2: VCP/T
    └── (integrated into signing services)

docs/
├── VCP_OVERVIEW.md          # Protocol specification
├── VCP_CONTEXT_DATA_FLOW.md # Dataflow reference
├── VCP_IMPLEMENTATION_GUIDE.md # Developer guide
├── identity/                # Layer 1 specs
├── semantics/               # Layer 3 specs
├── adaptation/              # Layer 4 specs
└── VCP_NEWCOMER_GUIDE.md    # This document

tests/vcp/                   # 195 tests
├── identity/
├── semantics/
├── adaptation/
└── integration/

Quick Start for Developers

1. Encode Context

from services.vcp import ContextEncoder

encoder = ContextEncoder()
ctx = encoder.encode(
    time="morning",
    space="home",
    company=["children"],
)
print(ctx.encode())  # "⏰🌅|📍🏡|👥👶"

2. Track State

from services.vcp import StateTracker

tracker = StateTracker(max_history=100)
transition = tracker.record(ctx)
if transition and transition.is_significant:
    print(f"Significant change: {transition.severity}")

3. Parse CSM1

from services.vcp.semantics import CSM1Parser

parser = CSM1Parser()
result = parser.parse("N5+F+E")
print(result.persona)  # "NANNY"
print(result.adherence_level)  # 5
print(result.scopes)  # ["FAMILY", "EDUCATION"]

4. Validate Token

from services.vcp.identity import TokenValidator

validator = TokenValidator()
token = validator.parse("family.safe.guide@1.2.0")
print(token.canonical)  # "family.safe.guide"
print(token.uri)  # "creed://creed.space/family.safe.guide@1.2.0"

Running Tests

# All VCP tests
python3 -m pytest tests/vcp/ -v

# Specific layer
python3 -m pytest tests/vcp/adaptation/ -v
python3 -m pytest tests/vcp/identity/ -v
python3 -m pytest tests/vcp/semantics/ -v

# With coverage
python3 -m pytest tests/vcp/ --cov=services/vcp

Key Documentation

Document	Purpose
VCP_OVERVIEW.md	Full protocol specification
VCP_CONTEXT_DATA_FLOW.md	Dataflow and security model
VCP_IMPLEMENTATION_GUIDE.md	Developer reference
VCP_ADAPTATION.md	Enneagram Protocol spec
VCP_SEMANTICS_CSM1.md	CSM1 grammar spec

Research Context

VCP implements the theoretical framework from the research proposal:

Proposal Goal	VCP Implementation
"What3Words for ethics"	VCP/I tokens: `family.safe.guide`
"Compact safety rules at runtime"	VCP/S CSM1: `N5+F+E`
"Emoji-based symbolic encoding"	VCP/A Enneagram: `⏰🌅\|📍🏡\|👥👶`
"Auditable value exchange"	VCP/T signed bundles + audit logs
"Inter-agent negotiation"	VCP signals in PDP pipeline

FilesExpand file tree

VCP_NEWCOMER_GUIDE.md

Latest commit

History

VCP_NEWCOMER_GUIDE.md

File metadata and controls

VCP Newcomer's Guide

What is VCP?

The Six Layers

Mapping to Research Proposal

Layer 1: VCP-Identity (VCP/I)

Token Format

URI Scheme

Key Files

Layer 2: VCP-Transport (VCP/T)

Purpose

Signed Bundles

Verification Chain

Trust Anchors

Audit Trail

Proof of Adherence

Key Files

Export Formats

Layer 3: VCP-Semantics (VCP/S)

CSM1 Grammar (Constitutional Safety Minicode)

Personas

Adherence Levels

Key Files

Layer 4: VCP-Adaptation (VCP/A)

The Enneagram Protocol

Wire Format

State Tracking

Transition Severity

Key Files

Dataflow: End to End

Storage & Persistence

Redis Keys (VCP Context)

What IS Stored

What is NOT Stored

HTTP API Endpoints

Example: Token Validation

Example: Context Encoding

MCP Integration

Conflict Resolution (Constitution Composition)

Composition Rules

Conflict Types

CSM1 Composition

Implementation

Conflict Detection

Inter-Agent Messaging (Future)

Concept

Message Format (Draft)

Use Cases (Planned)

Research Questions

Creed Space Frontend Integration

User Journey

UI Components

Example: Constitution Card

Live Preview (Chat Pane)

Export VCP Block

Key Frontend Files

Feature Flags

Directory Structure

Quick Start for Developers

1. Encode Context

2. Track State

3. Parse CSM1

4. Validate Token

Running Tests

Key Documentation

Research Context