Synarc
upflame/Synarc
A production-grade always-on engineering cognition runtime that provides architecture awareness, impact tracing, deployment risk analysis, and adaptive code intelligence across Claude Code, Claude Web, Codex CLI, Cursor, Windsurf, and SKILL.md runtimes.
Instalação
npx convez-ai upflame/Synarc1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
Synarc — Engineering Intelligence Runtime

Synarc is an engineering intelligence runtime for AI-assisted software development. It provides structured change classification, risk assessment, intent verification, audit trails, and session continuity across all major AI coding agents. The runtime operates as a skill pack — no executables, no network calls, no external dependencies.
Overview
AI coding agents execute changes rapidly, but without structured governance, each session risks context fragmentation, unplanned scope expansion, and undetected contract violations. Synarc addresses this by interposing a deterministic governance layer between developer intent and agent execution.
The runtime is organized as a multi-skill architecture:
- Intent Contracts — formal commitments before execution: scope boundary, explicit promises, risk cap, and post-execution verification
- Change Classification — 12 WorkTypes across 7 dimensions with deterministic risk floors
- Verification Engine — post-execution diff analysis against contract promises
- Audit & Compliance — immutable audit trail with rollback-to-intent and regulatory export (EU AI Act, SOC2, HIPAA)
- Session Continuity — persistent session ledger across context resets
- Cognition Mesh — multi-role team collaboration with shared working memory
Features
| Capability | Description |
|---|---|
| Intent Contracts | Propose → Accept → Execute → Verify → Fulfill lifecycle for every change |
| Intent Templates | 11 per-WorkType contract templates with default promises and scope rules |
| Verification Engine | Scope checking, promise verification, risk delta, composite verdict matrix |
| Audit Trail | Immutable records per contract lifecycle event, rollback-to-intent protocol |
| Change Classification | 12 WorkTypes, 7 dimensions, deterministic risk floors per domain |
| Risk Assessment | 6-level risk ladder with hard floors for auth, payments, schema changes |
| Context Injection | COMPACT (50 tokens), STANDARD (200 tokens), FULL (500 tokens) levels |
| Session Tracking | Immutable ledger persisting across sessions and context resets |
| Quality Gates | Zero-tolerance enforcement per WorkType |
| Error Intelligence | 6-step protocol with persistent error memory |
| Cognition Mesh | Multi-role collaboration — up to 9 specialized roles per task |
Architecture
The runtime implements a 7-step pipeline executed on every interaction:
Classify → Inject → Execute → Log → Aggregate → Checkpoint → Emit
Each step is deterministic and runtime-agnostic — the same pipeline operates identically across Claude Code, Codex CLI, Cursor, Windsurf, OpenCode, Gemini CLI, Copilot, Cline, and any AGENTS.md-compatible runtime.
Cache Architecture
| Tier | Contents | Scope |
|---|---|---|
| 0 | Pack header (AGENTS.md, manifest.yaml) | Session |
| 1 | Core reasoning (synarc-core, cognition-layer, schemas) | Session (~60 KB) |
| 2 | Active domain skill (one of 56) | Task (~10 KB each) |
| 3 | Skill references | Lazy-loaded |
| 4 | Dynamic context (project files, tool outputs) | Never cached |
Scale Adaptation
| Scale | Threshold | Tracking Depth |
|---|---|---|
| NANO | Single file, single purpose | WorkType + risk only |
| MICRO | 2-10 files | CURRENT_STATE.md |
| SMALL | <5k LOC, 1-5 modules | Full brain directory |
| MEDIUM | 5k-50k LOC, team | Full ledger |
| LARGE | 50k-500k LOC, multi-service | Service-boundary tracking |
| ENTERPRISE | >500k LOC, regulated | Compliance audit trail |
Auto-detected. Zero configuration.
Quick Start
Prerequisites
Any AI coding runtime: Claude Code, Codex CLI, Cursor, Windsurf, OpenCode, Gemini CLI, Copilot, or Cline.
Installation by Runtime
| Runtime | Method |
|---|---|
| Claude Code | claude plugin marketplace add upflame/Synarc then claude plugin install synarc |
| Codex CLI | Copy synarc-universal/AGENTS.md to repo root |
| OpenCode | Copy synarc-universal/AGENTS.md to repo root or ~/.config/opencode/AGENTS.md |
| Cursor | Copy .cursor/rules/synarc-core.mdc to .cursor/rules/ |
| Windsurf | Copy shared/runtime-adapters/windsurf.md to .windsurfrules |
| Gemini CLI | Copy synarc-universal/AGENTS.md to GEMINI.md in repo root |
| Copilot | Append shared/runtime-adapters/copilot.md to .github/copilot-instructions.md |
| Cline | Copy synarc-universal/skills/* to .cline/skills/ |
Verify Installation
> what did we change?
── Session Ledger ──
[14:00] FEATURE | auth/router.ts (+12, -3) | MEDIUM | IN_SCOPE
[14:05] FIX | auth/middleware.ts (+5, -0) | LOW | IN_SCOPE
Aggregate: MEDIUM (stable)
─────────────────────
Available Commands
| Command | Response |
|---|---|
what did we change? | Full session ledger |
summarize this session | Cognitive summary |
is this safe to deploy? | Risk delta + explicit YES/NO |
what tests are missing? | All unfilled test gaps |
generate a snapshot | Brain snapshot entry |
full handoff | Agent handoff block + brain updates |
run quality gates | All gates PASS/FAIL report |
Project Structure
synarc-universal/
├── AGENTS.md # Activation entry point
├── package.json # npm package manifest
├── manifest.yaml # Universal manifest
├── scripts/
│ ├── install.js # CLI installer
│ └── install.ps1 # PowerShell installer
├── skills/ # 56 domain skills
│ ├── synarc-core/ # Core runtime (always active)
│ ├── architect/ # System design & trade-off analysis
│ ├── backend-engineer/ # Service architecture & API design
│ ├── security-engineer/ # Threat modeling & defense
│ ├── debug-engineer/ # Systematic debugging
│ └── ... # 51 additional domain skills
├── shared/
│ ├── schemas/ # JSON Schema definitions
│ ├── workflows/ # Canonical workflow definitions
│ ├── guardrails/ # Constitutional safety rules
│ ├── standards/ # Naming conventions, frontmatter spec
│ ├── runtime-adapters/ # Per-runtime compilation rules
│ ├── prompts/ # Fallback prompt tiers
│ └── checklists/ # Review checklists
├── docs/ # Documentation
└── security/ # OWASP mapping, adversarial scenarios
Documentation
| Category | Links |
|---|---|
| Getting Started | Installation · Architecture · Usage · Deployment · Migration |
| Workflows | Change Classification · Risk Assessment · Context Injection · Session Tracking · Quality Gates · Error Intelligence |
| Intent Contracts | Contract Lifecycle · Templates · Verification · Audit & Compliance |
| Schemas | Intent Contract · Intent Template · Verification Result · Audit Record · Ledger Entry · Risk Assessment · Guardrails |
Security & Compliance
| Guard | Status |
|---|---|
| Sandboxed execution | Enabled |
| No network access | Verified |
| No filesystem write outside /brain/ | Enforced |
| Deterministic activation | Validated |
| Safe fallbacks on protocol error | Configured |
| Hash-verified integrity (SHA-256) | Active |
| Tamper protection | Enabled |
Risk hard floors:
| Domain | Minimum Risk | Reasoning |
|---|---|---|
| Auth, billing, payments, security | HIGH | Revenue, access, or trust impact |
| Schema change (remove/rename) | CRITICAL | Data integrity + migration complexity |
| Environment variable rename | CRITICAL | All deployments affected |
| Public API response change | HIGH | All consumers must adapt |
| Network / IAM config | CRITICAL | Security boundary change |
| INCIDENT response | CRITICAL | Production emergency |
Regulatory mappings: EU AI Act, SOC2, HIPAA — export templates included.
Contributing
See CONTRIBUTING.md for guidelines on submitting issues, feature requests, and pull requests.
This project follows Semantic Versioning and maintains a CHANGELOG.
License
MIT — see LICENSE.
Built by UpFlame Labs.