Skip to harness content
Open technical reference map

Mechanism source reference

Valdris SDLC Harness — Mermaid Maps

Complete canonical source reference from docs/REPO_MERMAID_MAPS.md.

Canonical pathdocs/REPO_MERMAID_MAPS.mdRevision69bab1cInspect source

Valdris SDLC Harness — Mermaid Maps

These diagrams are the repo operating map, not a raw import graph.

A raw diagram with every file would turn into spaghetti. These maps keep the repo readable by grouping files by responsibility, lane, gate, artifact, and runtime boundary.

How to read the maps

TermMeaning
LaneType of work routed through the harness: bug, feature, cloud, security, incident, QA, etc.
StageStep inside the work: intake, route, design, implement, prove, handoff.
GateBlocking verification check: proof, Red Zone, smoke, finish-line, self-heal.
LayerProduction-readiness dimension: frontend, backend, DB, auth, cloud, observability, recovery, etc.
ArtifactEvidence file proving the stage/gate happened.

Mode note: these are Blueprint diagrams. A real Live Run requires connector events from Claude Code, Codex, Hermes, MCP/API/CLI emitters, or watched artifacts.


1. Whole repo operating map

Source image: Whole repo operating map

flowchart TB
  classDef front fill:#082f49,stroke:#38bdf8,color:#e0f2fe,stroke-width:2px
  classDef app fill:#111827,stroke:#60a5fa,color:#dbeafe,stroke-width:2px
  classDef docs fill:#172554,stroke:#818cf8,color:#e0e7ff,stroke-width:2px
  classDef gen fill:#422006,stroke:#f59e0b,color:#fffbeb,stroke-width:2px
  classDef runtime fill:#134e4a,stroke:#2dd4bf,color:#ccfbf1,stroke-width:2px
  classDef gate fill:#4c0519,stroke:#fb7185,color:#ffe4e6,stroke-width:2px
  classDef evidence fill:#052e16,stroke:#22c55e,color:#dcfce7,stroke-width:2px

  subgraph FD["Front doors"]
    README["README.md"]
    AGENTS["AGENTS.md"]
    CLAUDE["CLAUDE.md"]
  end

  subgraph DOCS["Operating docs and contracts"]
    ARCH["docs/ARCHITECTURE.md"]
    CONTRACT["docs/CONNECTOR_EVENT_CONTRACT.md"]
    LANES["docs/SDLC_LANE_TAXONOMY.md"]
    MODES["docs/MODES_BLUEPRINT_LIVE_REPLAY.md"]
    PROD["docs/PRODUCTION_READINESS_LAYER_PACK.md"]
    PROOF["docs/ENTERPRISE_PROOF_BANK.md"]
    OI["docs/OPERATING_INTELLIGENCE_LAYER.md"]
    TESTDAY["docs/TEST_DAY_ACCEPTANCE_GATES.md"]
    MAP["docs/HARNESS_REPO_MAP.md"]
  end

  subgraph APP["Visual control-plane app"]
    ROUTES["app/"]
    COMPONENTS["components/"]
    LIB["lib/"]
    API["app/api/runs/demo/events"]
  end

  subgraph COMMISSION["Commissioning generator"]
    COMMISSION_SCRIPT["scripts/commission-harness.mjs"]
    TEMPLATES["templates/"]
    GENERATED["generated harness pack"]
  end

  subgraph RUNTIME["Runtime connector layer"]
    BRIDGE["scripts/claude-code-bridge.mjs"]
    EMITTER["scripts/uash-emit-event.mjs"]
    SIM["scripts/simulate-agent-run.mjs"]
  end

  subgraph CODE_INTELLIGENCE["Repo intelligence and anchors"]
    SCAN["scripts/code-intelligence-scan.mjs"]
    GITNEXUS["graph/gitnexus.json"]
    GRAPH_GATE["scripts/code-intelligence-gate.mjs"]
    ANCHOR_GATE["scripts/anchor-gate.mjs"]
    GRAPH["graph/graph.json"]
    ANCHORS["design/anchors.json"]
  end

  subgraph GATES["Verification and finish-line gates"]
    VERIFY["scripts/verify-harness.mjs"]
    PROOF["proof/proof.json"]
    SMOKE["smoke/smoke_proof.json"]
    REDZONE["approvals/redzone.json"]
    SELFHEAL["self_heal/self_heal_report.md"]
  end

  subgraph RUNS["Run packets and evidence"]
    RUN_TEMPLATE["runs/_run-template/"]
    EVENTS["session/events.jsonl"]
    HANDOFF["handoff/final.md"]
  end

  FD --> DOCS
  FD --> COMMISSION_SCRIPT
  DOCS --> COMMISSION_SCRIPT
  TEMPLATES --> COMMISSION_SCRIPT
  COMMISSION_SCRIPT --> GENERATED

  GENERATED --> AGENTS
  GENERATED --> CLAUDE
  GENERATED --> RUN_TEMPLATE
  GENERATED --> RUNTIME

  EMITTER --> BRIDGE
  SIM --> BRIDGE
  BRIDGE --> EVENTS

  SCAN --> GRAPH
  SCAN --> ANCHORS
  GRAPH --> GRAPH_GATE
  ANCHORS --> ANCHOR_GATE

  CODE_INTELLIGENCE --> GATES
  EVENTS --> GATES
  VERIFY --> GATES

  GATES --> HANDOFF
  GATES --> SELFHEAL

  ROUTES --> COMPONENTS
  COMPONENTS --> LIB
  API --> APP
  EVENTS --> APP
  GRAPH --> APP
  GATES --> APP

  class README,AGENTS,CLAUDE front
  class ROUTES,COMPONENTS,LIB,API app
  class ARCH,CONTRACT,LANES,MODES,PROD,MAP docs
  class COMMISSION_SCRIPT,TEMPLATES,GENERATED gen
  class BRIDGE,EMITTER,SIM runtime
  class VERIFY,PROOF,SMOKE,REDZONE,SELFHEAL,GRAPH_GATE,ANCHOR_GATE gate
  class RUN_TEMPLATE,EVENTS,HANDOFF,GRAPH,ANCHORS evidence

2. Universal core vs project adapter

This is the public product shape: universal harness logic stays reusable, while each repo gets a generated adapter.

Source image: Universal core vs project adapter

flowchart LR
  classDef universal fill:#082f49,stroke:#38bdf8,color:#e0f2fe,stroke-width:2px
  classDef adapter fill:#422006,stroke:#f59e0b,color:#fffbeb,stroke-width:2px
  classDef repo fill:#052e16,stroke:#22c55e,color:#dcfce7,stroke-width:2px
  classDef gate fill:#4c0519,stroke:#fb7185,color:#ffe4e6,stroke-width:2px

  USER["team or operator"] --> INTERVIEW["commissioning interview"]
  REPO["target repository"] --> SCAN["GitNexus code-intelligence index"]

  subgraph CORE["Universal core"]
    SCHEMA["adapter schema"]
    LANES["lane taxonomy"]
    STAGES["canonical stage flow"]
    EVENTS["connector event contract"]
    GATES["proof and approval gates"]
    MONITOR["visual flow monitor"]
  end

  subgraph ADAPTER["Generated project adapter"]
    TRUTH["source-of-truth order"]
    COMMANDS["validation commands"]
    REDZONE["Red Zone owners"]
    ENABLED["enabled lanes"]
    STYLE["answer and handoff style"]
  end

  subgraph TARGET["Project-specific harness pack"]
    A["AGENTS.md"]
    C["CLAUDE.md"]
    CMD["Claude slash command"]
    CODEX["Codex runtime prompt"]
    RUNS["run packet template"]
  end

  INTERVIEW --> CORE
  SCAN --> CORE
  CORE --> ADAPTER
  ADAPTER --> TARGET
  TARGET --> AGENTRUN["Claude Code / Codex / Hermes run"]
  AGENTRUN --> GATES
  GATES --> HANDOFF["proof-backed handoff"]

  class SCHEMA,LANES,STAGES,EVENTS,GATES,MONITOR universal
  class TRUTH,COMMANDS,REDZONE,ENABLED,STYLE adapter
  class A,C,CMD,CODEX,RUNS,TARGET repo
  class GATES,HANDOFF gate

3. Request routing, skills, and work lanes

Valdris uses two related routing tools. Workflow skills choose how a phase is performed. Commissioned lanes load project context, owners, commands, and gate emphasis. The route binds both without treating them as the same concept.

Source image: Request routing and eight-skill map

Source image: Valdris work lane families

flowchart TB
  classDef route fill:#082f49,stroke:#38bdf8,color:#e0f2fe,stroke-width:2px
  classDef skill fill:#111827,stroke:#818cf8,color:#e5e7eb,stroke-width:1.5px
  classDef goal fill:#451a03,stroke:#f59e0b,color:#fef3c7,stroke-width:2px
  classDef assurance fill:#052e16,stroke:#22c55e,color:#dcfce7,stroke-width:2px

  REQUEST["Natural-language request"] --> INTAKE["Authorized intake"]
  INTAKE --> CLASSIFY["Workload classification"]
  CLASSIFY --> ROUTE["Immutable route"]

  subgraph SKILLS["Eight workflow skills"]
    INTAKE_SKILL["valdris-intake-route"]
    BUG["valdris-bug-rca"]
    FEATURE["valdris-feature-delivery"]
    ARCH["valdris-architecture-refactor"]
    SECURITY["valdris-security-audit"]
    PLATFORM["valdris-platform-release"]
    GENAI["valdris-genai-assurance"]
    HANDOFF["valdris-proof-handoff"]
  end

  ROUTE --> SKILLS
  SKILLS --> PHASES["Intake phase → delivery phase → proof-handoff phase"]
  PHASES --> GOAL["Durable goal: budgets, checkpoints, stop conditions, expected revision"]
  PHASES --> ASSURANCE["Layer 0 + applicable production domains + concerns + packs"]
  GOAL --> DONE["Evidence-backed handoff"]
  ASSURANCE --> DONE

  class REQUEST,INTAKE,CLASSIFY,ROUTE route
  class INTAKE_SKILL,BUG,FEATURE,ARCH,SECURITY,PLATFORM,GENAI,HANDOFF skill
  class PHASES,GOAL goal
  class ASSURANCE,DONE assurance

4. Connector event flow

The connector is what turns external agent work into harness-observable events.

Source image: Connector event flow

flowchart LR
  classDef agent fill:#111827,stroke:#60a5fa,color:#dbeafe,stroke-width:2px
  classDef bridge fill:#134e4a,stroke:#2dd4bf,color:#ccfbf1,stroke-width:2px
  classDef store fill:#052e16,stroke:#22c55e,color:#dcfce7,stroke-width:2px
  classDef reject fill:#4c0519,stroke:#fb7185,color:#ffe4e6,stroke-width:2px
  classDef app fill:#172554,stroke:#818cf8,color:#e0e7ff,stroke-width:2px

  subgraph AGENTS["External coding-agent runtimes"]
    CLAUDE["Claude Code"]
    CODEX["Codex"]
    HERMES["Hermes"]
    FUTURE["future runtime"]
  end

  AGENTS --> EMIT["CLI emitter / MCP / API / watched artifact"]
  EMIT --> BRIDGE["local connector bridge"]

  BRIDGE --> VALIDATE["strict event validation"]
  VALIDATE --> STORE["event ledger and run packet"]
  STORE --> GATEENGINE["gate engine"]
  STORE --> UI["visual run monitor"]

  VALIDATE -->|bad event| REJECT1["reject: unknown node or bad schema"]
  GATEENGINE -->|missing artifact| REJECT2["block: finish-line not satisfied"]
  GATEENGINE -->|agent approval| REJECT3["block: Red Zone needs human"]
  GATEENGINE -->|harness gap| SELFHEAL["self-heal artifact or PR"]
  GATEENGINE -->|passed| HANDOFF["proof-backed handoff"]

  class CLAUDE,CODEX,HERMES,FUTURE,AGENTS agent
  class EMIT,BRIDGE,VALIDATE bridge
  class STORE store
  class GATEENGINE,UI app
  class REJECT1,REJECT2,REJECT3,SELFHEAL reject
  class HANDOFF store

5. 13-layer production readiness pack

The 13 layers are not the whole harness. They are a production-readiness pack that activates when a task touches real product, infra, deploy, users, data, security, or reliability.

Source image: 13-layer production readiness pack

flowchart TB
  classDef product fill:#082f49,stroke:#38bdf8,color:#e0f2fe,stroke-width:2px
  classDef platform fill:#422006,stroke:#f59e0b,color:#fffbeb,stroke-width:2px
  classDef ops fill:#134e4a,stroke:#2dd4bf,color:#ccfbf1,stroke-width:2px
  classDef gate fill:#4c0519,stroke:#fb7185,color:#ffe4e6,stroke-width:2px

  RUN["production-impacting run"] --> PACK["production readiness layer pack"]

  subgraph PRODUCT["Product surface"]
    L1["01 Frontend"]
    L2["02 Backend / API"]
    L3["03 Database / storage"]
    L4["04 Auth / permissions / RLS"]
  end

  subgraph PLATFORM["Platform surface"]
    L5["05 Hosting / deployment"]
    L6["06 Cloud / compute"]
    L7["07 CI/CD / version control"]
    L8["08 Security"]
    L9["09 Rate limiting"]
    L10["10 Caching / CDN"]
    L11["11 Load balancing / scaling"]
  end

  subgraph OPS["Operational survival"]
    L12["12 Logs / observability"]
    L13["13 Availability / recovery / DR"]
  end

  PACK --> PRODUCT
  PACK --> PLATFORM
  PACK --> OPS

  PRODUCT --> STATUS["required / passed / failed / pending / skipped with reason"]
  PLATFORM --> STATUS
  OPS --> STATUS
  STATUS --> PROOF["production/layer-assessment.json"]
  PROOF --> FINISH["finish-line gate"]

  class L1,L2,L3,L4,PRODUCT product
  class L5,L6,L7,L8,L9,L10,L11,PLATFORM platform
  class L12,L13,OPS ops
  class STATUS,PROOF,FINISH gate

6. Generated harness pack

This is what npm run commission creates for a target repo.

Source image: Generated harness pack

flowchart TB
  classDef input fill:#082f49,stroke:#38bdf8,color:#e0f2fe,stroke-width:2px
  classDef generated fill:#052e16,stroke:#22c55e,color:#dcfce7,stroke-width:2px
  classDef docs fill:#172554,stroke:#818cf8,color:#e0e7ff,stroke-width:2px
  classDef script fill:#422006,stroke:#f59e0b,color:#fffbeb,stroke-width:2px

  REPO["target repo"] --> COMMISSION["commission-harness.mjs"]
  ANSWERS["human answers"] --> COMMISSION
  GRAPH["GitNexus index"] --> COMMISSION

  subgraph FRONT["Agent front doors"]
    A["AGENTS.md"]
    C["CLAUDE.md"]
    SLASH[".claude/commands/valdris-sdlc-harness.md"]
    CODEX["docs/Codex Runtime Prompt.md"]
  end

  subgraph CONFIG["Project adapter"]
    ADAPTER["project-adapter.json"]
    YAML["project.yaml"]
    MAP["00_MAP.md"]
    CONTEXT["CONTEXT.md"]
  end

  subgraph OPERATING_DOCS["Operating docs"]
    VALIDATION["docs/Validation Commands.md"]
    REDZONE["docs/Red Zone Rules.md"]
    PROD["docs/Production Readiness Layers.md"]
    CLOUD["docs/Cloud Platform Engineering.md"]
    QA["docs/QA and Live Smoke.md"]
    MODES["docs/Modes Blueprint Live Replay.md"]
    SELF["docs/Self-Healing Loop.md"]
  end

  subgraph RUNPACK["Run packet"]
    TEMPLATE["runs/_run-template/README.md"]
    REVIEW["commissioning-review.md"]
  end

  subgraph GENERATED_SCRIPTS["Generated helper scripts"]
    EMITTER["scripts/uash-emit-event.mjs"]
    GSCAN["scripts/code-intelligence-scan.mjs"]
  end

  COMMISSION --> FRONT
  COMMISSION --> CONFIG
  COMMISSION --> OPERATING_DOCS
  COMMISSION --> RUNPACK
  COMMISSION --> GENERATED_SCRIPTS

  class REPO,ANSWERS,GRAPH,COMMISSION input
  class A,C,SLASH,CODEX,ADAPTER,YAML,MAP,CONTEXT,TEMPLATE,REVIEW generated
  class VALIDATION,REDZONE,PROD,CLOUD,QA,MODES,SELF docs
  class EMITTER,GSCAN script

7. README rule of thumb

For the README, keep only one simple overview image/diagram and link here for the deeper maps.

Recommended README line:

For lane-by-lane and repo-level Mermaid diagrams, see [Repo Mermaid Maps](docs/REPO_MERMAID_MAPS.md).

That keeps the README readable while still giving technical users a full visual map.