Agent Index

Last updated: April 11, 2026

Agent Index: Persona Reference

The dLANDiscord Assistant ecosystem is powered by a diverse workforce of 30+ specialized agents. Each agent has a distinct persona, model configuration, and tool authorization level.

Agent Classification

Agents are categorized by their cognitive complexity and role:

  • Planner: Capable of multi-step reasoning, task decomposition, and delegation. (Max 40 iterations).
  • Standard: Specialized workers focused on specific domains (e.g., Sysadmin, Security). (Max 20 iterations).
  • Special: System-level agents responsible for maintenance, evolution, or governance (e.g., MemoryMaster, MrTool).
  • One-Shot: Rapid-response agents that provide direct answers without a tool-use loop (e.g., google25pro).

Master Reference Table

Agent ID Type Role Description
architect Planner Overseer Matrix-style overseer; provides ATC control over the entire agent system.
nikki Planner Personal Assistant Master orchestrator; Dan's high-level personal assistant.
oracle Planner Strategic Planner Wise advisor; sees the "critical path" and creates master plans.
sasha Standard Local Sysadmin S.A.S.H.A. (System & Security Health Analyst); dLAN infra lead.
sysafe Standard Business SRE Primary SRE and Security Analyst for Sysafe.co.uk.
nova Standard Hosting Analyst AI support for London-Hosting.com.
cryptotrader Standard Crypto Analyst Market sentiment and trade direction analyst.
robocop Standard Home Surveillance 24/7 security and threat recognition for home CCTV.
gilfoyle Standard GitLab Analyst Cynical super-coder; expert in Go, Python, and CI/CD.
memorymaster Special Memory Architect Cognitive architect for system reflection and distillation.
projectmanager Special Hive Governor Creator of project workspaces and swarm orchestrator.
mrtool Special Tool Architect Eccentric architect for dynamic system tool creation.
teammanager Special Team Orchestrator Assembles and coordinates ad-hoc agent teams for missions.
analyst One-Shot Data Analyst High-level generator of SITREPs and trend analysis.

Team Member Personas (Team-Scoped)

These personas are only available within team contexts and are hidden from general planner visibility:

Persona Role Description
researcher Information Specialist Gathers, analyzes, and synthesizes information for team missions.
auditor Security Analyst Identifies vulnerabilities, risks, and compliance issues.
reviewer Quality Assurance Reviews team outputs for accuracy, completeness, and quality.
contributor General Executor Executes assigned tasks, provides supporting work.

Model Configuration & Fallbacks

Model assignments are defined in the agent's prompt Markdown header. The system uses a "Tiered Fallback" logic:

  1. Primary Model: The model requested in the .md header (e.g., ollama/qwen3:30b).
  2. Per-Agent Fallback: The secondary model defined in the .md header.
  3. Global Fallback: If both fail, the system falls back to the "Free Models" tier (Minimax, Gemini Lite).

Tool Registry (High-Level)

Tools are authorized based on the agent's persona and the user's role.

  • Shared Tools: Available to most agents (e.g., websearch, notify_user).
  • Infrastructure Tools: Restricted to sasha, sysadmin, and architect.
  • Business Tools: Restricted to agents in the BusinessHive (CEO, Worker).
  • Project Tools: Restricted to agents in AntFarm (Queen, Worker, Scout).
  • Team Tools: Restricted to TeamManager (tm_* management) and team members (tm_* member tools).
  • Meta Tools: Restricted to mrtool and architect for system modification.

Guidance for AI Agents

  • Persona Integrity: When operating as a specific agent, strictly adhere to the persona definition in your prompt.
  • Delegation: If a task is outside your ID's specialized role, use the call_agent tool to delegate to the appropriate specialist in this index.

Cross-References