AGENTS: The Workforce Behind Worksona

AGENTS is the core collection of Worksona's conversational AI agents. Where other projects in the portfolio handle delivery (MCP servers, APIs, registries), AGENTS is the workforce itself — the defined personas, reasoning frameworks, and strategic capabilities that give Worksona its character.

What is it?

The collection contains two primary offerings built on a shared architecture:

The Worksona Simulator is a strategic consulting environment. It uses three collaborating personas — Quin (facilitator), David (expert consultant), and Client (organizational stakeholder) — to walk leadership teams through AI adoption planning. It applies Wardley Mapping, Ansoff Matrix, Impact/Effort, Five Whys, TRIZ, and Job-To-Be-Done frameworks to the user's specific organizational context and exports a structured session document that can be handed to a real consultant.

Mr. Ranedeer is a personalized tutoring system. It adapts to 10 depth levels, 6 learning styles, 5 communication styles, 5 tone options, and 5 reasoning frameworks — producing over 7,500 unique tutor configurations from a single agent definition.

All agents are plain-text .txt prompt files. No servers, no databases, no compiled code. They load into ChatGPT, Claude, or any compatible LLM by paste or file upload.

Why is it useful?

Generic AI assistants are stateless, frameworkless, and inconsistent across sessions. Ask a vanilla model to apply Wardley Mapping to your organization and you will get a plausible-sounding but unstructured response that forgets your context on the next turn.

Worksona agents are different in three ways.

Prompt-as-Program. Each agent uses a pseudo-code architecture with named [Functions], [Init] blocks, [BEGIN]/[END] delimiters, and explicit state variables. The agent narrates its own execution — "I am calling [generate_questions]" — which forces structured, sequential reasoning rather than free-form response generation.

[Init]
    [BEGIN]
        say Hey Now!
        <introduce Quin and David and the purpose of this simulator>
        execute <create questions for the client>
        <Store QUESTIONS and EXAMPLE ANSWERS in the variable: QUESTIONS>
        say PROCEED or type MENU for a list of options.
    [END]

Session state via named variables. Quin v2 stores every major output in a labelled variable: CONFIGURATION, QUESTIONS, FRAMEWORKS, SUMMARY, STRATEGY. These variables persist across conversation turns. At any point, the user can call /export and receive a complete structured document covering the full session — importable by a consultant or shareable with a team.

Frameworks as first-class agent skills. David does not look up Wardley Mapping when asked. It is embedded in his definition as a named function. This is an early implementation of what the AI field now calls skill libraries — capabilities baked into an agent's identity rather than injected per-session.

How and where does it apply?

The agent persona format uses a shared schema across all agents in the collection:

[Configuration]
  Name: Quin the Questionnator
  Role: Facilitator and success agent
  Depth: Master's level
  Communication Style: Socratic
  Tone Style: Encouraging
  Reasoning Framework: Inductive

[Functions]
  [generate_questions]
    [BEGIN]
      <analyze CONFIGURATION for concerns>
      <generate targeted questions>
      <store in QUESTIONS>
    [END]

[Commands]
  /menu:   Execute <show_menu>
  /export: Execute <export_session>

This schema makes every agent portable and composable. Swapping the client.txt persona changes the organizational context without touching Quin's logic. Adding a new strategic framework to David means adding one named function block to his file. New personas follow the same template and are immediately compatible with the orchestrating agent.

The reference table in the repository lists 17 additional specialist agents — The Concept Architect, The System Cartographer, The Strategy Weaver, The Framework Architect, and others — each defined with a three-layer analysis model and a distinct reasoning approach. These extend the workforce beyond consulting simulation into strategic analysis, knowledge architecture, trend interpretation, and process optimization.

AGENTS sits at the bottom of the Worksona stack and at the top of its value delivery. Everything else — registries, MCP servers, APIs — exists to route work to these definitions. The workforce is the product.