memex/notes/org-skill-model-explorer.org

SKILL: Model Explorer Agent (Universal Literate Note)

• Overview
• Phase A: Demand (PRD)
  • 1. Purpose
  • 2. User Needs
  • 3. Success Criteria
    • Command Trigger Verification (@agent list models)
    • Provider Registry Introspection
    • Org Table Formatting Accuracy
• Phase B: Blueprint (PROTOCOL)
  • 1. Architectural Intent
  • 2. Semantic Interfaces
• Phase D: Build (Implementation)
  • Provider Introspection
• Registration

Overview

The Model Explorer Agent provides dynamic introspection of the system's LLM capabilities. It intercepts specific user commands to list and describe all available models across providers, rendering them as native Org-mode tables.

Phase A: Demand (PRD)

1. Purpose

Define the interfaces for system-wide model discovery and transparency.

2. User Needs

• Transparency: Visible list of models and context windows.
• Determinism: Metadata retrieval must bypass System 1 for high fidelity.
• Integration: Results rendered as native Org-mode tables.

3. Success Criteria

TODO Command Trigger Verification (@agent list models)

TODO Provider Registry Introspection

TODO Org Table Formatting Accuracy

Phase B: Blueprint (PROTOCOL)

1. Architectural Intent

Interfaces for dynamic skill introspection and Emacs UI injection. Source of truth is the active `*skills-registry*`.

2. Semantic Interfaces

(defun trigger-skill-model-explorer (context)
  "Triggers on '@agent list models' in buffer updates.")

(defun build-org-table-for-models ()
  "Dynamically builds an Org table from registered provider skills.")

(defun execute-skill-model-explorer (proposed-action context)
  "Injects the model table into the active Emacs buffer.")

Phase D: Build (Implementation)

Provider Introspection

(defun build-org-table-for-models ()
  (let ((table-rows (list "| Provider | Model ID | Context |" "|----------+----------+---------|")))
    ;; Logic to iterate through skills-registry and call get-available-models
    (format nil "~{~a~^~%~}" table-rows)))

Registration

(defskill :skill-model-explorer
  :priority 85
  :trigger #'trigger-skill-model-explorer
  :neuro (lambda (context) nil)
  :symbolic #'execute-skill-model-explorer)