# purpleailab/decepticon architecture

> Architecture diagram of the GitHub repository purpleailab/decepticon, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-04-16). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/purpleailab/decepticon
- Repository: https://github.com/purpleailab/decepticon

## Overview

1. **Overall shape**
   Decepticon is split into a **Python control plane** and a **TypeScript terminal client**. The Python side contains the autonomous agent runtime, tool integrations, planning, and state/observability. The CLI is an Ink-based TUI that streams progress, renders messages, and exposes engagement controls. The repo is designed around **two isolated Docker networks**: a management plane and an operational sandbox plane.

2. **Operator entrypoint**
   The main user-facing entrypoint is the CLI under `clients/cli`. It provides the REPL, keyboard shortcuts, status panels, and a local graph/canvas view. It is not just a shell wrapper; it is the operator console for starting, stopping, and monitoring engagements while the backend executes objectives.

3. **Engagement planning subsystem**
   `decepticon/agents/soundwave.py` and the related `skills/soundwave/*` content generate the mission package: **ConOps, Rules of Engagement, Deconfliction Plan, and OPPLAN**. This is the planning boundary in the system. The OPPLAN becomes the structured input to the autonomous loop, while the RoE acts as a guardrail throughout execution.

4. **Autonomous orchestration loop**
   The core control logic lives in `decepticon/core/engagement.py`, `engagement_loop.py`, and `orchestrator.py`. The loop selects the next pending objective, assembles a prompt from the objective plus prior findings and constraints, spawns a fresh agent context, and interprets the result as success, blocked, or pending. State is persisted outside the model context so objectives do not depend on long-lived agent memory.

5. **Specialized agent model**
   The repository defines multiple specialist agents under `decepticon/agents/` such as recon, exploit, post-exploit, analyst, defender, patcher, and research-oriented roles. Each agent has its own prompt material in `decepticon/agents/prompts/` and is intended to run with a clean context window. This creates a clear boundary between planning, execution, analysis, and reporting responsibilities.

6. **Skills as on-disk knowledge packs**
   The `skills/` tree is a major subsystem. Skills are organized by domain and kill-chain phase, and they are loaded progressively rather than all at once. This gives the agents a curated knowledge base for AD, cloud, web, contracts, reversing, post-exploit, and shared workflow/OPSEC material, with MITRE ATT&CK metadata embedded where relevant.

7. **Tool execution layer**
   `decepticon/tools/` contains the concrete action surface used by agents: bash command execution, web, cloud, Active Directory, contracts, reversing, research, defense, and reporting tools. These tools are what turn a model’s plan into measurable actions and artifacts. The split by domain is important for the architecture diagram because it shows where capabilities are isolated and how different agent roles access different toolsets.

8. **Sandbox and interactive sessions**
   Execution is intentionally separated from management. `decepticon/backends/docker_sandbox.py` and the container assets under `containers/` point to a hardened operational sandbox, with interactive shell session handling and prompt detection built for long-lived tools like C2 clients or shells. The README’s Sliver profile indicates an optional **C2 subsystem** running in the operational network, with the sandbox using preconfigured client access to interact with it.

9. **LLM access and routing**
   The `decepticon/llm/` package and `config/litellm.yaml` form the model-access boundary. A LiteLLM proxy routes requests to different providers/models, and the router/factory layer selects models by role with fallback behavior. This means the orchestrator does not talk to a single model backend directly; it goes through a configurable abstraction that supports profile-based model choice.

10. **Knowledge, graphs, and findings**
   `decepticon/tools/research/`, `decepticon/tools/references/`, and `decepticon/schemas/` support structured knowledge and evidence handling. The presence of `neo4j_store.py`, attack-graph docs, and graph-related CLI code suggests a graph-backed view of findings, relationships, and attack paths. Findings are also surfaced in reporting tools, so the data flow is: tool output → structured state/graph → CLI/report output.

11. **Guardrails, observability, and safety**
   `decepticon/middleware/` enforces operational constraints such as OPPLAN hierarchy, safe command execution, and skill exposure rules. `decepticon/observability/` adds activity logging, metrics, and tracing, which is important because the system is autonomous and needs an audit trail. This is the boundary that keeps the agent loop inspectable and policy-aware.

12. **Build, runtime, and deployment**
   The repo is designed to run via **Docker Compose**, with separate images for the CLI, sandbox, LangGraph-related orchestration, and optional C2 components. The Python package is the backend runtime, while the TypeScript client is the operator interface. CI, tests, and container definitions are all present, which indicates the system is meant to be exercised as an integrated multi-service platform rather than as a single library.

Main entry points: [`decepticon/agents/decepticon.py`](https://github.com/purpleailab/decepticon/blob/main/decepticon/agents/decepticon.py) (Specialist Agents), [`decepticon/core/engagement_loop.py`](https://github.com/purpleailab/decepticon/blob/main/decepticon/core/engagement_loop.py) (Engagement Loop), [`clients/cli/src/screens/REPL.tsx`](https://github.com/purpleailab/decepticon/blob/main/clients/cli/src/screens/REPL.tsx) (CLI REPL), [`decepticon/tools/research/tools.py`](https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/research/tools.py) (Research Tools)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_group_client["TypeScript CLI"]
  node_node_cli_repl["CLI REPL<br/>terminal UI<br/>[REPL.tsx]"]
  node_node_cli_state["UI State<br/>app state<br/>[store.ts]"]
  node_node_cli_canvas["Graph Canvas<br/>canvas view<br/>[graphView.ts]"]
  node_node_cli_messages["Message Views<br/>rendering<br/>[AIMessage.tsx]"]
end

subgraph group_group_control["Python control plane"]
  node_node_orchestrator["Orchestrator<br/>entrypoint<br/>[orchestrator.py]"]
  node_node_engagement["Engagement Loop<br/>runtime loop<br/>[engagement_loop.py]"]
  node_node_engagement_core["Engagement Core<br/>state machine<br/>[engagement.py]"]
  node_node_agent_pool["Specialist Agents<br/>agent runtime<br/>[decepticon.py]"]
  node_node_soundwave["Mission Planning<br/>planning boundary<br/>[soundwave.py]"]
  node_node_middleware["Guardrails<br/>policy middleware<br/>[opplan.py]"]
end

subgraph group_group_skills["Skills library"]
  node_node_skills_soundwave["Planning Skills<br/>skill packs<br/>[SKILL.md]"]
  node_node_skills_recon["Recon Skills<br/>skill packs<br/>[SKILL.md]"]
  node_node_skills_exploit["Exploit Skills<br/>skill packs<br/>[SKILL.md]"]
  node_node_skills_post["Post-Exploit Skills<br/>skill packs<br/>[SKILL.md]"]
  node_node_skills_shared["Shared Skills<br/>skill packs<br/>[SKILL.md]"]
end

subgraph group_group_tools["Tool surface"]
  node_node_tools_recon["Research Tools<br/>tool domain<br/>[tools.py]"]
  node_node_tools_web["Web Tools<br/>tool domain<br/>[tools.py]"]
  node_node_tools_ad["AD Tools<br/>tool domain<br/>[tools.py]"]
  node_node_tools_cloud["Cloud Tools<br/>tool domain<br/>[tools.py]"]
  node_node_tools_contracts["Contract Tools<br/>tool domain<br/>[tools.py]"]
  node_node_tools_bash["Bash Tools<br/>execution tool<br/>[bash.py]"]
end

subgraph group_group_runtime["Runtime & deployment"]
  node_node_llm["LLM Router<br/>model access<br/>[router.py]"]
  node_node_sandbox["Sandbox Backend<br/>container backend<br/>[docker_sandbox.py]"]
  node_node_containers["Container Stack<br/>docker assets<br/>[sandbox.Dockerfile]"]
end

subgraph group_group_data["State & evidence"]
  node_node_opplan["Plan Schema<br/>structured plan<br/>[schemas.py]"]
  node_node_obs["Observability<br/>audit trail<br/>[activity.py]"]
  node_node_research_graph["Evidence Graph<br/>knowledge graph<br/>[graph.py]"]
  node_node_reporting["Reporting<br/>report generation<br/>[executive.py]"]
end

node_node_orchestrator -->|"starts"| node_node_engagement
node_node_engagement -->|"drives"| node_node_engagement_core
node_node_engagement -->|"plans"| node_node_soundwave
node_node_engagement -->|"spawns"| node_node_agent_pool
node_node_engagement -->|"routes prompts"| node_node_llm
node_node_engagement -->|"constrained by"| node_node_middleware
node_node_agent_pool -->|"loads"| node_node_skills_recon
node_node_agent_pool -->|"loads"| node_node_skills_exploit
node_node_agent_pool -->|"loads"| node_node_skills_post
node_node_agent_pool -->|"loads"| node_node_skills_shared
node_node_soundwave -->|"uses"| node_node_skills_soundwave
node_node_middleware -->|"validates"| node_node_opplan
node_node_agent_pool -->|"executes"| node_node_tools_bash
node_node_agent_pool -->|"uses"| node_node_tools_recon
node_node_agent_pool -->|"uses"| node_node_tools_web
node_node_agent_pool -->|"uses"| node_node_tools_ad
node_node_agent_pool -->|"uses"| node_node_tools_cloud
node_node_agent_pool -->|"uses"| node_node_tools_contracts
node_node_tools_recon -->|"writes"| node_node_research_graph
node_node_tools_recon -->|"summarizes"| node_node_reporting
node_node_research_graph -->|"feeds"| node_node_reporting
node_node_obs -->|"observes"| node_node_engagement
node_node_obs -->|"observes"| node_node_agent_pool
node_node_cli_repl -->|"updates"| node_node_cli_state
node_node_cli_repl -->|"renders"| node_node_cli_messages
node_node_cli_repl -->|"shows"| node_node_cli_canvas
node_node_cli_repl -->|"controls"| node_node_orchestrator
node_node_cli_canvas -->|"visualizes"| node_node_research_graph
node_node_sandbox -->|"hosts"| node_node_tools_bash
node_node_containers -->|"packages"| node_node_sandbox
node_node_containers -->|"packages"| node_node_cli_repl
node_node_llm -->|"consumes"| node_node_opplan

click node_node_cli_repl "https://github.com/purpleailab/decepticon/blob/main/clients/cli/src/screens/REPL.tsx"
click node_node_cli_state "https://github.com/purpleailab/decepticon/blob/main/clients/cli/src/state/store.ts"
click node_node_cli_canvas "https://github.com/purpleailab/decepticon/blob/main/clients/cli/src/canvas/graphView.ts"
click node_node_cli_messages "https://github.com/purpleailab/decepticon/blob/main/clients/cli/src/components/messages/AIMessage.tsx"
click node_node_orchestrator "https://github.com/purpleailab/decepticon/blob/main/decepticon/orchestrator.py"
click node_node_engagement "https://github.com/purpleailab/decepticon/blob/main/decepticon/core/engagement_loop.py"
click node_node_engagement_core "https://github.com/purpleailab/decepticon/blob/main/decepticon/core/engagement.py"
click node_node_agent_pool "https://github.com/purpleailab/decepticon/blob/main/decepticon/agents/decepticon.py"
click node_node_soundwave "https://github.com/purpleailab/decepticon/blob/main/decepticon/agents/soundwave.py"
click node_node_middleware "https://github.com/purpleailab/decepticon/blob/main/decepticon/middleware/opplan.py"
click node_node_llm "https://github.com/purpleailab/decepticon/blob/main/decepticon/llm/router.py"
click node_node_sandbox "https://github.com/purpleailab/decepticon/blob/main/decepticon/backends/docker_sandbox.py"
click node_node_containers "https://github.com/purpleailab/decepticon/blob/main/containers/sandbox.Dockerfile"
click node_node_opplan "https://github.com/purpleailab/decepticon/blob/main/decepticon/core/schemas.py"
click node_node_obs "https://github.com/purpleailab/decepticon/blob/main/decepticon/observability/activity.py"
click node_node_research_graph "https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/research/graph.py"
click node_node_reporting "https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/reporting/executive.py"
click node_node_skills_soundwave "https://github.com/purpleailab/decepticon/blob/main/skills/soundwave/conops-template/SKILL.md"
click node_node_skills_recon "https://github.com/purpleailab/decepticon/blob/main/skills/recon/active-recon/SKILL.md"
click node_node_skills_exploit "https://github.com/purpleailab/decepticon/blob/main/skills/exploit/web/SKILL.md"
click node_node_skills_post "https://github.com/purpleailab/decepticon/blob/main/skills/post-exploit/c2/SKILL.md"
click node_node_skills_shared "https://github.com/purpleailab/decepticon/blob/main/skills/shared/workflow/SKILL.md"
click node_node_tools_recon "https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/research/tools.py"
click node_node_tools_web "https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/web/tools.py"
click node_node_tools_ad "https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/ad/tools.py"
click node_node_tools_cloud "https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/cloud/tools.py"
click node_node_tools_contracts "https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/contracts/tools.py"
click node_node_tools_bash "https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/bash/bash.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_node_cli_repl,node_node_cli_state,node_node_cli_canvas,node_node_cli_messages toneBlue
class node_node_orchestrator,node_node_engagement,node_node_engagement_core,node_node_agent_pool,node_node_soundwave,node_node_middleware toneAmber
class node_node_skills_soundwave,node_node_skills_recon,node_node_skills_exploit,node_node_skills_post,node_node_skills_shared toneMint
class node_node_tools_recon,node_node_tools_web,node_node_tools_ad,node_node_tools_cloud,node_node_tools_contracts,node_node_tools_bash toneRose
class node_node_llm,node_node_sandbox,node_node_containers toneIndigo
class node_node_opplan,node_node_obs,node_node_research_graph,node_node_reporting toneTeal
```

## Components

### TypeScript CLI

Operator-facing terminal client and local canvas UI.

- [CLI REPL](https://github.com/purpleailab/decepticon/blob/main/clients/cli/src/screens/REPL.tsx): terminal UI · Ink-based operator console, commands, keybindings, and engagement controls. · `clients/cli/src/screens/REPL.tsx`
- [UI State](https://github.com/purpleailab/decepticon/blob/main/clients/cli/src/state/store.ts): app state · Client-side store, hooks, and session state for rendering engagement progress. · `clients/cli/src/state/store.ts`
- [Graph Canvas](https://github.com/purpleailab/decepticon/blob/main/clients/cli/src/canvas/graphView.ts): canvas view · Local graph/canvas rendering for objectives, events, and subagent activity. · `clients/cli/src/canvas/graphView.ts`
- [Message Views](https://github.com/purpleailab/decepticon/blob/main/clients/cli/src/components/messages/AIMessage.tsx): rendering · Typed message and status components for tool calls, agents, and reports. · `clients/cli/src/components/messages/AIMessage.tsx`

### Python control plane

Autonomous engagement runtime, agents, middleware, and orchestration.

- [Orchestrator](https://github.com/purpleailab/decepticon/blob/main/decepticon/orchestrator.py): entrypoint · Process entrypoint that wires the runtime together. · `decepticon/orchestrator.py`
- [Engagement Loop](https://github.com/purpleailab/decepticon/blob/main/decepticon/core/engagement_loop.py): runtime loop · Selects objectives, builds prompts, interprets results, and persists state. · `decepticon/core/engagement_loop.py`
- [Engagement Core](https://github.com/purpleailab/decepticon/blob/main/decepticon/core/engagement.py): state machine · Core engagement coordination, objective tracking, and lifecycle management. · `decepticon/core/engagement.py`
- [Specialist Agents](https://github.com/purpleailab/decepticon/blob/main/decepticon/agents/decepticon.py): agent runtime · Role-specific agents for recon, exploit, post-exploit, analysis, defense, and research. · `decepticon/agents/decepticon.py`
- [Mission Planning](https://github.com/purpleailab/decepticon/blob/main/decepticon/agents/soundwave.py): planning boundary · Builds ConOps, Rules of Engagement, deconfliction, and OPPLAN inputs. · `decepticon/agents/soundwave.py`
- [Guardrails](https://github.com/purpleailab/decepticon/blob/main/decepticon/middleware/opplan.py): policy middleware · OPPLAN hierarchy, safe command execution, and skill exposure rules. · `decepticon/middleware/opplan.py`

### Skills library

On-disk knowledge packs and mission templates loaded by agents.

- [Planning Skills](https://github.com/purpleailab/decepticon/blob/main/skills/soundwave/conops-template/SKILL.md): skill packs · Mission templates for threat profile, ConOps, RoE, and OPPLAN conversion. · `skills/soundwave/conops-template/SKILL.md`
- [Recon Skills](https://github.com/purpleailab/decepticon/blob/main/skills/recon/active-recon/SKILL.md): skill packs · Passive, active, cloud, OSINT, and web recon knowledge packs. · `skills/recon/active-recon/SKILL.md`
- [Exploit Skills](https://github.com/purpleailab/decepticon/blob/main/skills/exploit/web/SKILL.md): skill packs · Web, AD, and general exploit workflows and references. · `skills/exploit/web/SKILL.md`
- [Post-Exploit Skills](https://github.com/purpleailab/decepticon/blob/main/skills/post-exploit/c2/SKILL.md): skill packs · Credential access, lateral movement, privilege escalation, and C2 guidance. · `skills/post-exploit/c2/SKILL.md`
- [Shared Skills](https://github.com/purpleailab/decepticon/blob/main/skills/shared/workflow/SKILL.md): skill packs · Workflow, OPSEC, defense-evasion, and finding-protocol material shared by agents. · `skills/shared/workflow/SKILL.md`

### Tool surface

Domain-specific action tools used by agents to execute work.

- [Research Tools](https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/research/tools.py): tool domain · Research, CVE, graph, fuzzing, patching, and scanner utilities. · `decepticon/tools/research/tools.py`
- [Web Tools](https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/web/tools.py): tool domain · HTTP, GraphQL, OAuth, JWT, and session-oriented web tooling. · `decepticon/tools/web/tools.py`
- [AD Tools](https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/ad/tools.py): tool domain · Active Directory enumeration, Kerberos, DCSync, and BloodHound-oriented actions. · `decepticon/tools/ad/tools.py`
- [Cloud Tools](https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/cloud/tools.py): tool domain · AWS, Kubernetes, metadata, and Terraform-oriented cloud actions. · `decepticon/tools/cloud/tools.py`
- [Contract Tools](https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/contracts/tools.py): tool domain · Smart contract analysis, patterns, Foundry, and Slither helpers. · `decepticon/tools/contracts/tools.py`
- [Bash Tools](https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/bash/bash.py): execution tool · Shell command execution and prompt handling for local or sandboxed actions. · `decepticon/tools/bash/bash.py`

### Runtime & deployment

Model routing, sandboxing, containers, and compose-level runtime wiring.

- [LLM Router](https://github.com/purpleailab/decepticon/blob/main/decepticon/llm/router.py): model access · Role-based model selection, provider routing, and fallback handling. · `decepticon/llm/router.py`
- [Sandbox Backend](https://github.com/purpleailab/decepticon/blob/main/decepticon/backends/docker_sandbox.py): container backend · Operational sandbox execution and interactive session management. · `decepticon/backends/docker_sandbox.py`
- [Container Stack](https://github.com/purpleailab/decepticon/blob/main/containers/sandbox.Dockerfile): docker assets · Dockerfiles and entrypoints for CLI, sandbox, orchestration, and optional C2. · `containers/sandbox.Dockerfile`

### State & evidence

Structured findings, graphs, reporting, and observability.

- [Plan Schema](https://github.com/purpleailab/decepticon/blob/main/decepticon/core/schemas.py): structured plan · Structured engagement plan and finding schemas used across the runtime. · `decepticon/core/schemas.py`
- [Observability](https://github.com/purpleailab/decepticon/blob/main/decepticon/observability/activity.py): audit trail · Activity logging, metrics, and tracing for autonomous execution. · `decepticon/observability/activity.py`
- [Evidence Graph](https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/research/graph.py): knowledge graph · Structured research state, Neo4j-backed graphing, and relationship storage. · `decepticon/tools/research/graph.py`
- [Reporting](https://github.com/purpleailab/decepticon/blob/main/decepticon/tools/reporting/executive.py): report generation · Executive, timeline, and platform-specific report outputs. · `decepticon/tools/reporting/executive.py`

## Connections

- Orchestrator → Engagement Loop: starts · Bootstraps the engagement runtime.
- Engagement Loop → Engagement Core: drives · Loop delegates lifecycle and objective coordination.
- Engagement Loop → Mission Planning: plans · Consumes mission package and OPPLAN inputs.
- Engagement Loop → Specialist Agents: spawns · Creates fresh specialist contexts per objective.
- Engagement Loop → LLM Router: routes prompts · Uses model routing instead of direct provider access.
- Engagement Loop → Guardrails: constrained by · Execution and skill exposure are policy-checked.
- Specialist Agents → Recon Skills: loads · Recon role pulls recon skills and references.
- Specialist Agents → Exploit Skills: loads · Exploit-oriented roles use exploit skills.
- Specialist Agents → Post-Exploit Skills: loads · Post-exploit roles use post-exploit skills.
- Specialist Agents → Shared Skills: loads · All roles share workflow and OPSEC guidance.
- Mission Planning → Planning Skills: uses · Planning templates and rules come from soundwave skills.
- Guardrails → Plan Schema: validates · Enforces plan hierarchy and objective constraints.
- Specialist Agents → Bash Tools: executes · Agents can invoke shell execution through the tool layer.
- Specialist Agents → Research Tools: uses · Research and triage roles access research tooling.
- Specialist Agents → Web Tools: uses · Web-focused agents use HTTP and auth tooling.
- Specialist Agents → AD Tools: uses · AD roles use directory and Kerberos tooling.
- Specialist Agents → Cloud Tools: uses · Cloud roles use cloud inspection and access tooling.
- Specialist Agents → Contract Tools: uses · Contract roles use contract analysis tooling.
- Research Tools → Evidence Graph: writes · Research outputs feed graph-backed evidence state.
- Research Tools → Reporting: summarizes · Findings are converted into reports and briefs.
- Evidence Graph → Reporting: feeds · Graph-backed findings drive reporting artifacts.
- Observability → Engagement Loop: observes · Logs, metrics, and traces cover the autonomous loop.
- Observability → Specialist Agents: observes · Captures agent activity and subagent streaming.
- CLI REPL → UI State: updates · UI renders from local client state and hooks.
- CLI REPL → Message Views: renders · Shows agent, tool, and system messages.
- CLI REPL → Graph Canvas: shows · Operator can inspect graph/canvas views.
- CLI REPL → Orchestrator: controls · CLI starts, stops, and monitors engagements.
- Graph Canvas → Evidence Graph: visualizes · Canvas presents relationships and findings.
- Sandbox Backend → Bash Tools: hosts · Interactive execution happens inside sandboxed sessions.
- Container Stack → Sandbox Backend: packages · Container assets define the sandbox runtime.
- Container Stack → CLI REPL: packages · CLI container builds the operator console.
- LLM Router → Plan Schema: consumes · Model prompts are shaped by the structured plan.
