- Added Agent Lightning research section to researcher.html with Demo 2 results - Created comprehensive /integrations/agent-lightning.html page - Added Agent Lightning link in homepage hero section - Updated Discord invite links (Tractatus + semantipy) across all pages - Added feedback.js script to all key pages for live demonstration Phase 2 of Master Plan complete: Discord setup → Website completion 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
632 lines
18 KiB
Markdown
632 lines
18 KiB
Markdown
# Governance Authorization System - Architecture Plan
|
|
|
|
**Classification**: INTERNAL
|
|
**Date**: November 2, 2025
|
|
**Status**: Planning Phase
|
|
**Security Level**: Not for public distribution
|
|
|
|
## Executive Summary
|
|
|
|
This document outlines the architecture for a scalable authorization system to protect the integrity of the Tractatus Governance Framework. The current system can be bypassed via scripts - this is intentional for legitimate operations but requires proper authorization controls.
|
|
|
|
**Core Principle**: Authorization complexity scales with organizational complexity.
|
|
|
|
## Problem Statement
|
|
|
|
### Current State
|
|
|
|
The governance framework protects instruction-history.json via PreToolUse hooks, preventing direct modifications through Claude Code's Edit/Write tools. However, as demonstrated in session 2025-11-02:
|
|
|
|
```bash
|
|
# This bypasses the governance hook
|
|
cat > .claude/instruction-history.json << 'EOF'
|
|
{ "modified": "content" }
|
|
EOF
|
|
```
|
|
|
|
**Why This Exists**: Legitimate rule updates must be possible (e.g., adding security rules). The hook protects against *accidental* violations, not *authorized* changes.
|
|
|
|
**The Gap**: No authorization mechanism distinguishes legitimate from malicious bypasses.
|
|
|
|
### Security Requirements
|
|
|
|
1. **Authenticity**: Verify the actor is authorized to modify rules
|
|
2. **Authorization**: Verify the actor has permission for *this specific change*
|
|
3. **Audit Trail**: Record who approved what, when, and why
|
|
4. **Non-Repudiation**: Cryptographic proof of approval
|
|
5. **Scalability**: Single developer → multinational organization
|
|
6. **Least Privilege**: Minimum necessary access for each role
|
|
|
|
## Proposed Architecture
|
|
|
|
### 1. Vault Integration
|
|
|
|
**Purpose**: Secure credential and token storage
|
|
|
|
**Technology Options**:
|
|
- HashiCorp Vault (enterprise-grade)
|
|
- AWS Secrets Manager (cloud-native)
|
|
- Azure Key Vault (Microsoft ecosystem)
|
|
- Self-hosted vault (tractatus-integrated)
|
|
|
|
**Vault Structure**:
|
|
```
|
|
vault/
|
|
├── credentials/
|
|
│ ├── developers/
|
|
│ │ ├── developer1/
|
|
│ │ │ ├── totp_secret
|
|
│ │ │ ├── pgp_public_key
|
|
│ │ │ └── authorization_level
|
|
│ │ └── developer2/...
|
|
│ ├── approvers/
|
|
│ │ ├── team_leads/...
|
|
│ │ ├── department_heads/...
|
|
│ │ └── executives/...
|
|
├── tokens/
|
|
│ ├── active/
|
|
│ │ └── token_abc123_expiry_timestamp
|
|
│ └── revoked/
|
|
│ └── audit_trail
|
|
└── policies/
|
|
├── rule_modification_requirements.json
|
|
└── approval_chains.json
|
|
```
|
|
|
|
### 2. Authorization Tiers
|
|
|
|
**Tier 0: Single Developer (2FA)**
|
|
- Scenario: Personal projects, small startups
|
|
- Requirement: Time-based OTP (TOTP) or hardware key (YubiKey)
|
|
- Approval: Self-approval with 2FA verification
|
|
- Example: Family-history project current state
|
|
|
|
**Tier 1: Small Team (Team Lead Approval)**
|
|
- Scenario: 2-10 developers
|
|
- Requirement: Developer requests, team lead approves
|
|
- Approval: Team lead signs with PGP key
|
|
- Example: Startup with tech lead
|
|
|
|
**Tier 2: Department (Management Approval)**
|
|
- Scenario: 10-50 developers across departments
|
|
- Requirement: Team lead requests, department head approves
|
|
- Approval: Department head + compliance review
|
|
- Example: Mid-size company
|
|
|
|
**Tier 3: Enterprise (Multi-Level Approval)**
|
|
- Scenario: 50-500 developers
|
|
- Requirement: Department → VP → CTO approval chain
|
|
- Approval: Digital signatures at each level
|
|
- Example: Large corporation
|
|
|
|
**Tier 4: Multinational (Board/Compliance Approval)**
|
|
- Scenario: 500+ developers across countries
|
|
- Requirement: Regional compliance → Corporate compliance → Legal → Board
|
|
- Approval: Complex approval matrix based on rule category
|
|
- Example: Global enterprise with regulatory requirements
|
|
|
|
### 3. Rule Security Classification
|
|
|
|
**Current State**: All rules treated equally
|
|
|
|
**Proposed**: Tag each rule with security classification
|
|
|
|
```json
|
|
{
|
|
"id": "inst_fh_sec_001",
|
|
"title": "Client-Side Storage Prohibition",
|
|
"securityClassification": "INTERNAL",
|
|
"modificationRequiresApproval": "TIER_1",
|
|
"publicDisclosure": "SUMMARY_ONLY",
|
|
"rationale": "Full implementation details could expose attack surface"
|
|
}
|
|
```
|
|
|
|
**Classification Levels**:
|
|
|
|
1. **PUBLIC**: Safe for public documentation
|
|
- General principles (e.g., "Multi-tenant isolation required")
|
|
- No implementation details
|
|
- Broad security goals
|
|
|
|
2. **INTERNAL**: Company employees only
|
|
- Implementation patterns
|
|
- Technology choices
|
|
- Non-sensitive technical details
|
|
|
|
3. **CONFIDENTIAL**: Need-to-know basis
|
|
- Specific security controls
|
|
- Vulnerability mitigations
|
|
- Integration details
|
|
|
|
4. **RESTRICTED**: Executive/Compliance only
|
|
- Vault integration details
|
|
- Cryptographic key management
|
|
- Incident response procedures
|
|
|
|
5. **SECRET**:極秘 (Highest classification)
|
|
- Master keys
|
|
- Emergency override procedures
|
|
- Security vulnerability details before patches
|
|
|
|
### 4. Authorization Workflow
|
|
|
|
#### Scenario: Developer Needs to Add Security Rule
|
|
|
|
**Step 1: Request Initiation**
|
|
```bash
|
|
# Developer runs authorization request tool
|
|
./scripts/governance/request-rule-modification.sh \
|
|
--action add \
|
|
--rule-id inst_fh_sec_014 \
|
|
--justification "Add XSS prevention rule per security audit" \
|
|
--urgency normal
|
|
|
|
# Output:
|
|
# Request ID: REQ-2025-11-02-001
|
|
# Required Approval: TIER_1 (Team Lead)
|
|
# Token will be generated upon approval
|
|
```
|
|
|
|
**Step 2: Approval Chain**
|
|
```bash
|
|
# Team lead reviews request
|
|
./scripts/governance/review-request.sh --request-id REQ-2025-11-02-001
|
|
|
|
# Shows:
|
|
# - Proposed rule changes (diff)
|
|
# - Justification
|
|
# - Risk assessment
|
|
# - Related rules affected
|
|
|
|
# Team lead approves with PGP signature
|
|
./scripts/governance/approve-request.sh \
|
|
--request-id REQ-2025-11-02-001 \
|
|
--pgp-sign \
|
|
--comments "Approved - aligns with security policy"
|
|
```
|
|
|
|
**Step 3: Token Generation**
|
|
```bash
|
|
# System generates time-limited authorization token
|
|
# Token stored in vault with:
|
|
# - Request ID
|
|
# - Approver signature
|
|
# - Expiry timestamp (e.g., 1 hour)
|
|
# - Allowed operation (add rule inst_fh_sec_014 only)
|
|
|
|
# Developer receives token
|
|
# Token: GOVAUTH_REQ2025110201_EXP20251102T1200_SIG_abc123def456
|
|
```
|
|
|
|
**Step 4: Authorized Modification**
|
|
```bash
|
|
# Developer uses token to modify rules
|
|
./scripts/governance/add-rule-authorized.sh \
|
|
--token GOVAUTH_REQ2025110201_EXP20251102T1200_SIG_abc123def456 \
|
|
--rule-file new-rule-inst_fh_sec_014.json
|
|
|
|
# Script validates:
|
|
# 1. Token signature (PGP verify)
|
|
# 2. Token not expired
|
|
# 3. Token not revoked
|
|
# 4. Operation matches token scope
|
|
# 5. Vault confirms token validity
|
|
|
|
# If valid, modifies instruction-history.json
|
|
# Logs to GovernanceAuditLog with approval chain
|
|
```
|
|
|
|
**Step 5: Audit Trail**
|
|
```javascript
|
|
// GovernanceAuditLog entry
|
|
{
|
|
sessionId: "session-2025-11-02-xyz",
|
|
timestamp: "2025-11-02T11:45:00Z",
|
|
service: "AuthorizedRuleModification",
|
|
operation: "add_rule",
|
|
ruleId: "inst_fh_sec_014",
|
|
requestId: "REQ-2025-11-02-001",
|
|
requester: "developer@company.com",
|
|
approvers: [
|
|
{
|
|
name: "teamlead@company.com",
|
|
role: "TIER_1_APPROVER",
|
|
signature: "PGP_SIGNATURE_HASH",
|
|
timestamp: "2025-11-02T11:30:00Z"
|
|
}
|
|
],
|
|
tokenUsed: "GOVAUTH_REQ2025110201_...",
|
|
justification: "Add XSS prevention rule per security audit",
|
|
evidence: {
|
|
beforeHash: "sha256_of_instruction_history_before",
|
|
afterHash: "sha256_of_instruction_history_after",
|
|
diff: "git diff output"
|
|
}
|
|
}
|
|
```
|
|
|
|
### 5. Emergency Override System
|
|
|
|
**Purpose**: Critical security fixes when normal approval chain too slow
|
|
|
|
**Requirements**:
|
|
- Break-glass mechanism
|
|
- Multi-signature required (2 of 3 executives)
|
|
- Automatic incident report
|
|
- Post-incident review mandatory
|
|
|
|
**Process**:
|
|
```bash
|
|
# Emergency override request
|
|
./scripts/governance/emergency-override.sh \
|
|
--severity CRITICAL \
|
|
--vulnerability CVE-2025-XXXXX \
|
|
--justification "Zero-day exploit in production" \
|
|
--rule-id inst_fh_sec_015
|
|
|
|
# Requires 2 of 3 executive signatures within 15 minutes
|
|
# If approved, generates 15-minute token
|
|
# Automatically creates incident report
|
|
# Schedules post-incident review meeting
|
|
```
|
|
|
|
### 6. Approval Chain Configuration
|
|
|
|
**File**: `.claude/governance/approval-chains.json`
|
|
|
|
```json
|
|
{
|
|
"organizationTier": "TIER_1",
|
|
"approvalChains": {
|
|
"ruleModification": {
|
|
"add": {
|
|
"securityClassification": {
|
|
"PUBLIC": ["SELF_2FA"],
|
|
"INTERNAL": ["TEAM_LEAD"],
|
|
"CONFIDENTIAL": ["TEAM_LEAD", "DEPARTMENT_HEAD"],
|
|
"RESTRICTED": ["DEPARTMENT_HEAD", "CTO", "COMPLIANCE"],
|
|
"SECRET": ["CTO", "COMPLIANCE", "CEO", "LEGAL"]
|
|
}
|
|
},
|
|
"modify": {
|
|
"securityClassification": {
|
|
"PUBLIC": ["TEAM_LEAD"],
|
|
"INTERNAL": ["TEAM_LEAD", "DEPARTMENT_HEAD"],
|
|
"CONFIDENTIAL": ["DEPARTMENT_HEAD", "CTO"],
|
|
"RESTRICTED": ["CTO", "COMPLIANCE", "LEGAL"],
|
|
"SECRET": ["BOARD_APPROVAL"]
|
|
}
|
|
},
|
|
"delete": {
|
|
"securityClassification": {
|
|
"PUBLIC": ["TEAM_LEAD", "COMPLIANCE"],
|
|
"INTERNAL": ["DEPARTMENT_HEAD", "COMPLIANCE"],
|
|
"CONFIDENTIAL": ["CTO", "COMPLIANCE"],
|
|
"RESTRICTED": ["CTO", "COMPLIANCE", "LEGAL"],
|
|
"SECRET": ["BOARD_APPROVAL", "LEGAL"]
|
|
}
|
|
}
|
|
},
|
|
"emergencyOverride": {
|
|
"required": ["EXECUTIVE_1", "EXECUTIVE_2"],
|
|
"outOf": 3,
|
|
"timeWindow": "15_MINUTES",
|
|
"postIncidentReview": "MANDATORY"
|
|
}
|
|
},
|
|
"approvers": {
|
|
"TEAM_LEAD": ["teamlead@company.com"],
|
|
"DEPARTMENT_HEAD": ["depthead@company.com"],
|
|
"CTO": ["cto@company.com"],
|
|
"COMPLIANCE": ["compliance@company.com"],
|
|
"LEGAL": ["legal@company.com"],
|
|
"CEO": ["ceo@company.com"],
|
|
"EXECUTIVE_1": ["exec1@company.com"],
|
|
"EXECUTIVE_2": ["exec2@company.com"],
|
|
"EXECUTIVE_3": ["exec3@company.com"]
|
|
},
|
|
"vault": {
|
|
"provider": "hashicorp",
|
|
"endpoint": "https://vault.company.internal:8200",
|
|
"authMethod": "ldap",
|
|
"tokenTTL": "1h",
|
|
"maxTokenUses": 1
|
|
}
|
|
}
|
|
```
|
|
|
|
### 7. Documentation Security Tiers
|
|
|
|
**Problem**: Some documentation should never be public
|
|
|
|
**Solution**: Multi-tier documentation system
|
|
|
|
**Directory Structure**:
|
|
```
|
|
docs/
|
|
├── public/ # Safe for GitHub, website
|
|
│ ├── README.md
|
|
│ ├── GETTING_STARTED.md
|
|
│ └── governance/
|
|
│ └── PRINCIPLES.md # Broad principles only
|
|
├── internal/ # Company employees only
|
|
│ ├── IMPLEMENTATION.md
|
|
│ ├── ARCHITECTURE.md
|
|
│ └── governance/
|
|
│ └── RULE_CATALOG.md
|
|
├── confidential/ # Need-to-know (not in git)
|
|
│ ├── VAULT_INTEGRATION.md
|
|
│ ├── APPROVAL_WORKFLOWS.md
|
|
│ └── governance/
|
|
│ └── SECURITY_CONTROLS.md
|
|
└── restricted/ # Executive/Compliance (vault-stored)
|
|
├── EMERGENCY_PROCEDURES.md
|
|
├── INCIDENT_RESPONSE.md
|
|
└── governance/
|
|
└── MASTER_KEY_MANAGEMENT.md
|
|
```
|
|
|
|
**Git Configuration**:
|
|
```gitignore
|
|
# .gitignore
|
|
docs/confidential/
|
|
docs/restricted/
|
|
.claude/governance/approval-chains.json # Contains real emails/roles
|
|
.claude/governance/vault-config.json
|
|
```
|
|
|
|
**Vault Storage**: Confidential and Restricted docs stored in vault, retrieved only by authorized users
|
|
|
|
### 8. Implementation Phases
|
|
|
|
#### Phase 1: Foundation (Weeks 1-2)
|
|
- [ ] Design vault schema
|
|
- [ ] Create authorization token system
|
|
- [ ] Implement PGP signature verification
|
|
- [ ] Add securityClassification field to schema v3.0
|
|
- [ ] Create approval-chains.json template
|
|
|
|
#### Phase 2: Core Authorization (Weeks 3-4)
|
|
- [ ] Implement request-rule-modification.sh
|
|
- [ ] Implement review-request.sh
|
|
- [ ] Implement approve-request.sh
|
|
- [ ] Implement add-rule-authorized.sh
|
|
- [ ] Integrate with GovernanceAuditLog
|
|
|
|
#### Phase 3: Vault Integration (Weeks 5-6)
|
|
- [ ] Set up HashiCorp Vault (or alternative)
|
|
- [ ] Configure LDAP/OAuth integration
|
|
- [ ] Implement token generation/validation
|
|
- [ ] Create vault CLI wrapper scripts
|
|
- [ ] Migrate existing credentials to vault
|
|
|
|
#### Phase 4: Multi-Tier Support (Weeks 7-8)
|
|
- [ ] Implement approval chain engine
|
|
- [ ] Support TIER_0 through TIER_4
|
|
- [ ] Add emergency override system
|
|
- [ ] Create admin dashboard for approval management
|
|
- [ ] Testing across all tiers
|
|
|
|
#### Phase 5: Documentation Security (Weeks 9-10)
|
|
- [ ] Categorize existing documentation
|
|
- [ ] Move confidential docs to vault
|
|
- [ ] Create public-safe versions of sensitive docs
|
|
- [ ] Implement documentation access controls
|
|
- [ ] Update platform-admin hub to respect classification
|
|
|
|
### 9. Technology Stack
|
|
|
|
**Vault**: HashiCorp Vault
|
|
- Industry standard
|
|
- Excellent API
|
|
- Dynamic secrets support
|
|
- Audit logging built-in
|
|
|
|
**Signatures**: GnuPG (PGP)
|
|
- Cryptographic non-repudiation
|
|
- Widely adopted
|
|
- Offline signing support
|
|
- Air-gap capable for highest security
|
|
|
|
**2FA**: TOTP (Time-based OTP)
|
|
- Standards-based (RFC 6238)
|
|
- Works offline
|
|
- Compatible with Google Authenticator, Authy, etc.
|
|
- No phone/SMS dependency
|
|
|
|
**Audit**: MongoDB GovernanceAuditLog
|
|
- Already in use
|
|
- Supports complex queries
|
|
- Immutable append-only with proper configuration
|
|
- Easy integration with existing framework
|
|
|
|
**Token Format**: JWT (JSON Web Tokens)
|
|
- Self-contained
|
|
- Cryptographically signed
|
|
- Expiry built-in
|
|
- Industry standard
|
|
|
|
### 10. Security Considerations
|
|
|
|
**Attack Vectors to Mitigate**:
|
|
|
|
1. **Token Theft**:
|
|
- Mitigation: Short TTL (1 hour), single-use tokens
|
|
|
|
2. **Replay Attacks**:
|
|
- Mitigation: Nonce in token, vault tracks used tokens
|
|
|
|
3. **Man-in-Middle**:
|
|
- Mitigation: TLS for vault communication, PGP signatures
|
|
|
|
4. **Insider Threat**:
|
|
- Mitigation: Multi-party approval, audit trails, least privilege
|
|
|
|
5. **Compromised Approver**:
|
|
- Mitigation: Multi-signature for sensitive operations
|
|
|
|
6. **Social Engineering**:
|
|
- Mitigation: Out-of-band verification for high-risk changes
|
|
|
|
7. **Script Injection**:
|
|
- Mitigation: Validate all inputs, sanitize rule content
|
|
|
|
8. **Vault Compromise**:
|
|
- Mitigation: Hardware security module (HSM) for master keys
|
|
|
|
**Defense in Depth**:
|
|
- Multiple approval levels
|
|
- Cryptographic signatures
|
|
- Audit logging
|
|
- Time-limited tokens
|
|
- Single-use tokens
|
|
- Anomaly detection
|
|
- Incident response procedures
|
|
|
|
### 11. Public Documentation Guidelines
|
|
|
|
**What to Include in Public Docs**:
|
|
- ✅ General governance principles
|
|
- ✅ Rule categories and purposes
|
|
- ✅ Benefits of framework
|
|
- ✅ How to get started
|
|
- ✅ Example rules (non-sensitive)
|
|
|
|
**What to EXCLUDE from Public Docs**:
|
|
- ❌ Vault integration details
|
|
- ❌ Authorization token formats
|
|
- ❌ Approval chain specifics
|
|
- ❌ Emergency override procedures
|
|
- ❌ Specific security controls
|
|
- ❌ Cryptographic key details
|
|
- ❌ Script bypass techniques
|
|
|
|
**Public Documentation Template**:
|
|
```markdown
|
|
# Governance Framework Authorization
|
|
|
|
The Tractatus Framework includes enterprise-grade authorization
|
|
controls for rule modifications. The system scales from individual
|
|
developers to multinational organizations.
|
|
|
|
## Key Features
|
|
|
|
- **Multi-tier approval**: Configurable approval chains
|
|
- **Cryptographic verification**: Non-repudiation of approvals
|
|
- **Audit trails**: Complete record of all changes
|
|
- **Vault integration**: Secure credential storage
|
|
|
|
## Getting Started
|
|
|
|
Contact your system administrator to configure authorization
|
|
for your organization.
|
|
|
|
For implementation details, see internal documentation.
|
|
```
|
|
|
|
### 12. Migration Path for Existing Projects
|
|
|
|
**Current State**: No authorization (bypass via scripts possible)
|
|
|
|
**Target State**: Full vault-based authorization
|
|
|
|
**Migration Steps**:
|
|
|
|
1. **Phase 0: Audit** (Week 1)
|
|
- Identify all scripts that modify instruction-history.json
|
|
- Document current access patterns
|
|
- Assess organization tier (TIER_0 through TIER_4)
|
|
|
|
2. **Phase 1: Soft Enforcement** (Weeks 2-3)
|
|
- Add authorization checks to scripts (warn, don't block)
|
|
- Educate developers on new process
|
|
- Set up vault infrastructure
|
|
|
|
3. **Phase 2: Hard Enforcement** (Week 4)
|
|
- Convert warnings to errors
|
|
- Require authorization tokens
|
|
- Grandfather existing rules
|
|
|
|
4. **Phase 3: Full Compliance** (Week 5+)
|
|
- All modifications require authorization
|
|
- Regular audit reviews
|
|
- Continuous monitoring
|
|
|
|
### 13. Metrics and Monitoring
|
|
|
|
**Key Metrics**:
|
|
- Authorization request response time
|
|
- Approval chain completion rate
|
|
- Token expiry without use (indicates friction)
|
|
- Emergency override frequency
|
|
- Audit log coverage
|
|
|
|
**Alerts**:
|
|
- Unauthorized modification attempt
|
|
- Token reuse detected
|
|
- Approval chain timeout
|
|
- Emergency override used
|
|
- Vault connectivity issues
|
|
|
|
**Dashboard Widgets** (for platform-admin hub):
|
|
- Pending authorization requests
|
|
- Recent rule modifications
|
|
- Approval chain health
|
|
- Security classification distribution
|
|
- Token usage patterns
|
|
|
|
## Recommendations
|
|
|
|
### For Family-History Project (Current)
|
|
|
|
**Recommended Tier**: TIER_0 (Single Developer with 2FA)
|
|
|
|
**Immediate Actions**:
|
|
1. Set up TOTP for John G Stroh
|
|
2. Create simple 2FA wrapper for rule modification scripts
|
|
3. Add securityClassification field to existing rules
|
|
4. Move to TIER_1 if additional developers join
|
|
|
|
### For Platform-Admin Hub
|
|
|
|
**Recommended Features**:
|
|
1. Authorization request dashboard
|
|
2. Pending approvals view
|
|
3. Audit trail visualization
|
|
4. Documentation classification viewer
|
|
|
|
### For Tractatus Framework
|
|
|
|
**Enhancements Needed**:
|
|
1. Add securityClassification to schema v3.0
|
|
2. Create authorization subsystem
|
|
3. Vault integration module
|
|
4. Public documentation sanitization tools
|
|
|
|
## Conclusion
|
|
|
|
This authorization system provides:
|
|
- **Scalability**: Single developer → multinational org
|
|
- **Security**: Multiple layers of protection
|
|
- **Auditability**: Complete non-repudiable trail
|
|
- **Flexibility**: Configurable approval chains
|
|
- **Compliance**: Meets enterprise requirements
|
|
|
|
The system maintains the balance between:
|
|
- **Accessibility**: Legitimate changes possible
|
|
- **Security**: Unauthorized changes prevented
|
|
- **Usability**: Not overly burdensome
|
|
- **Transparency**: Public disclosure carefully managed
|
|
|
|
**Next Steps**: Begin Phase 1 implementation with vault setup and token system.
|
|
|
|
---
|
|
|
|
**Classification**: INTERNAL
|
|
**Author**: Claude (Anthropic) under direction of John G Stroh
|
|
**Review Required**: Before any public disclosure
|
|
**Related Documents**:
|
|
- `docs/GOVERNANCE_FRAMEWORK.md` (public version)
|
|
- `docs/SCHEMA_V3_SPECIFICATION.md`
|
|
- `.claude/instruction-history.json`
|