Skip to content

nddipiazza/robos / .agents/skills/manage-agent-personas

manage-agent-personas

Verified

Add, customize, configure, or inspect RobOS Developer Agent Personas (such as Non-Headless Developer, Human+Agent Desktop Co-Pilot, Software Architect, etc.) and their directives, execution environments, and token-saving IDE refactoring suites.

Manage RobOS Agent Personas & Personalities

Configure and customize RobOS Developer Agent Personas (robos:AgentPersona). RobOS equips AI agents (Claude Code, GitHub Copilot, Google Antigravity, Oh My Pi) with first-class specialized engineering roles rather than generic coding loops.

Input

$ARGUMENTS — Action and persona parameters:

  • list — List all registered built-in and custom developer personas
  • show <slug|id> — Inspect a specific persona definition, system prompt, and guidance
  • add <slug> "<role>" "<description>" [executionMode] — Scaffold a new developer persona
  • customize <slug> — Guide updating development guidance, system prompts, or refactoring rules
  • reset — Restore all personas to RobOS built-in defaults

Persona Architecture & Storage

RobOS manages agent personas across two synchronization tiers:

  1. Built-in System Personas: Defined in packages/robos-lib/agent-personas.js. These are deployed to /usr/local/share/robos/robos-lib/agent-personas.js and serve as canonical out-of-the-box defaults:

    • urn:robos:agent:software-architect — C4 domain modeling, ADRs, KGraph blast radius
    • urn:robos:agent:frontend-web-dev — Electron preload security, React 18, RobOS dark theme CSS
    • urn:robos:agent:game-dev — Godot 4 LTS, GDScript, isometric cRPGs, D&D 5e SRD
    • urn:robos:agent:backend-dev — Java 21 Spring Boot 3 / Node.js Fastify, OpenAPI 3.1, gRPC
    • urn:robos:agent:data-engineer-dev — PostgreSQL ACID schemas, Flyway migrations, Kafka streaming
    • urn:robos:agent:devops-engineer — Multi-cluster Kubernetes, Helm charts, ArgoCD GitOps
    • urn:robos:agent:non-headless-dev — Ephemeral Xvfb GUI developer, mapped IDE via IPC, step debugger, Chrome DevTools MCP, and token-saving IDE refactorings
    • urn:robos:agent:human-agent-copilot — Non-headless desktop co-pilot running in user's active session (DISPLAY=:0), collaborative IDE pairing, shared step debugging
  2. User Custom Personas & Overrides: Stored persistently in ~/.config/robos/agent-personas.json. When present, entries override built-in personas with matching IDs or add brand-new specialized roles.

  3. In-App Management: Both Task Implementer and Task Planner provide the interactive 🤖 Agent Personas & Prompts modal for reviewing, editing, creating, and resetting personas live.


Procedures

1. Adding a New Agent Persona

Option A: Adding a Built-in Persona (Repository Level)

  1. Open packages/robos-lib/agent-personas.js.
  2. Add an entry to BUILTIN_AGENT_PERSONAS:
    {
      id: 'urn:robos:agent:<slug>',
      slug: '<slug>',
      role: '<Display Role Name>',
      title: 'RobOS <Display Role Name> Agent',
      description: '<Role description and focus>',
      icon: '<Emoji Icon>',
      category: '<Category>',
      executionMode: '<headless|ephemeral-gui|desktop-session>',
      systemPrompt: `You are a specialized <Role> in RobOS...`,
      developmentGuidance: `1. Rule 1...\n2. Rule 2...`,
      planningPrompt: `Break down the task...`,
      implementationPrompt: `Implement the solution...`,
      tags: ['tag1', 'tag2'],
      modelPreference: 'pro',
      isDefault: true,
    }
    
  3. Update fallback lists in packages/task-implementer/renderer/app.js and packages/task-planner/renderer/app.js.
  4. Add detection keywords to detectPersonaForTask() in packages/robos-lib/agent-personas.js.

Option B: Adding a Custom Persona via File / Script

Create or append to ~/.config/robos/agent-personas.json:

[
  {
    "id": "urn:robos:agent:embedded-firmware-dev",
    "slug": "embedded-firmware-dev",
    "role": "Embedded Firmware Developer",
    "title": "RobOS Embedded Firmware Developer Agent",
    "description": "Specialist in Rust no_std, Zephyr RTOS, and hardware register simulation.",
    "icon": "⚡",
    "category": "Embedded Systems",
    "executionMode": "headless",
    "systemPrompt": "You are an Embedded Systems Firmware Developer in RobOS.",
    "developmentGuidance": "1. Memory Safety: Prefer Rust no_std.\n2. Concurrency: Use RTIC or Zephyr mutexes without busy-waiting.\n3. Hardware Emulation: Test drivers using Renode or QEMU ARM.",
    "planningPrompt": "Analyze hardware peripherals, memory maps, and interrupt priorities.",
    "implementationPrompt": "Write register abstraction layers, HAL implementations, and unit tests.",
    "tags": ["embedded", "rust", "zephyr", "firmware"],
    "isDefault": false
  }
]

Option C: Adding via Task Implementer / Task Planner GUI

  1. Open RobOS Task Implementer or RobOS Task Planner.
  2. Click 🤖 Agent Personas in the header.
  3. Click + Add Custom Persona in the sidebar.
  4. Fill in the Role, Title, Description, System Prompt, and Development Guidance.
  5. Click 💾 Save Persona.

2. Execution Modes & Environment Configuration

Execution ModeTarget DisplayWorkspace StoragePrimary Use Case
headlessN/A (CLI)Standard git clone or repo directoryPure code generation, CLI test runs, background builds
ephemeral-guiEphemeral Xvfb (:99)Ephemeral tmpfs / /tmp/robos-workspaces/Non-Headless Developer: Full GUI IDE execution, simulated user inputs via plugins, step debugger, Chrome DevTools MCP, zero host screen pollution
desktop-sessionHost display (:0)User's active development workspaceHuman + Agent Desktop Co-Pilot: Collaborative pairing in the user's active IDE, shared terminal, live breakpoint stepping, synchronized refactorings

3. Token-Saving IDE Refactorings Suite

When defining directives for GUI developers (non-headless-dev and human-agent-copilot), always instruct the agent to utilize IDE AST refactorings instead of full-file LLM rewrites:

TOKEN-SAVING IDE REFACTORING DIRECTIVE:
NEVER re-generate entire source files or massive diff blocks when making structural changes.
Always invoke RobOS IDE Bridge MCP refactoring tools:
- `robos_ide_refactor_rename`: Rename classes, methods, fields, and local variables across the AST without rewriting files.
- `robos_ide_refactor_extract_method`: Extract repetitive or complex blocks into dedicated methods.
- `robos_ide_refactor_extract_variable`: Extract expressions into local variables.
- `robos_ide_refactor_move`: Relocate classes/modules to target packages with automated import updating.
- `robos_ide_refactor_safe_delete`: Safely verify and prune unused symbols.
- `robos_ide_execute_action`: Run IDE native actions (ReformatCode, OptimizeImports).
This drastically lowers token consumption, avoids context window truncation, and eliminates hallucinated syntax errors.

4. Interactive Step Debugging & Chrome DevTools MCP

Instruct the agent to use developer tools rather than guessing:

  1. Step-by-Step Debugger (ide-bridge-mcp):

    • robos_ide_set_breakpoint: Set line breakpoints in target source files.
    • robos_ide_run_config: Launch the application or test in debug mode.
    • robos_ide_get_thread_state: Inspect the paused call stack frames and local variables.
    • robos_ide_register_breakpoint_webhook: Receive automated callbacks when a breakpoint hits.
  2. Chrome DevTools MCP (chrome-devtools-mcp):

    • take_snapshot / evaluate_script: Inspect live DOM and evaluate JavaScript.
    • list_console_messages: Inspect client warnings and uncaught exceptions.
    • list_network_requests: Inspect API request payloads and HTTP status codes.

5. Verification & Testing

Verify that personas and detection heuristics function accurately:

# 1. Test Task Planner Personas & Fallback Catalog
node --test packages/robos-test/tests/task-planner/agent-personas.test.js

# 2. Test Task Implementer Personas, Dropdown & Execution Directives
node --test packages/robos-test/tests/task-implementer/agent-personas.test.js

# 3. Test IDE Bridge MCP Refactoring Suite
node --test packages/robos-test/tests/mcp-servers/ide-bridge-mcp.test.js
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 470kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall