Onboarding Identity Map

Onboarding Identity Map

This document defines where onboarding data belongs, why it belongs there, and how A.I.M. should treat each file during ingestion.

Design Rules

1. GEMINI.md or AGENTS.md lives at the repo root. It is the active operating prompt for the coding agent.

2. Core identity belongs in core/. This includes structured facts about the operator, durable profile information, and machine-readable runtime configuration.

3. memory/ and continuity/ are not identity stores. They support memory refinement, recall, checkpoints, proposals, and continuity artifacts produced over time.

4. synapse/ is for external expert knowledge and ingest-only reference material. It should not be the primary home for the operator's own identity.

File Roles

GEMINI.md or AGENTS.md

Purpose: The live instruction surface for the agent.

What belongs here:

• agent identity
• operator name
• execution style defaults
• retrieval mandates
• planning rules
• guardrails

What does not belong here:

• detailed operator biography
• long-form social/personality analysis
• machine configuration
• evolving memory summaries

Why: This file should stay prompt-efficient and behavior-oriented.

core/OPERATOR.md

Purpose: Canonical structured operator identity.

What belongs here:

• legal or preferred name
• tech stack
• working style
• stable personal principles
• mission or goals
• business context
• concise factual profile fields gathered during onboarding

What does not belong here:

• shell paths
• model routing
• API endpoints
• verbose persona prose
• rolling memory summaries

Why: OPERATOR.md should be the stable source of truth for factual operator metadata.

core/OPERATOR_PROFILE.md

Purpose: Narrative operator profile used to capture richer persona and behavioral inference.

What belongs here:

• Grok-derived profile text
• communication style summary
• philosophical and social archetype notes
• inferred problem-solving style
• voice and tone guidance

What does not belong here:

• hard configuration
• filesystem paths
• API keys
• generated session memory

Why: This file complements OPERATOR.md. OPERATOR.md is structured fact. OPERATOR_PROFILE.md is richer interpretive context.

core/MEMORY.md

Purpose: Durable top-level memory anchor for A.I.M.

What belongs here:

• concise long-term state
• durable project posture
• high-signal operator/project continuity notes

What does not belong here:

• raw onboarding questionnaire dumps
• low-level runtime config
• temporary chat residue

Why: MEMORY.md is a compact continuity document, not the canonical operator profile.

core/CONFIG.json

Purpose: Machine-readable runtime configuration.

What belongs here:

• path configuration
• provider/model defaults
• service endpoints
• operational settings
• allowed root
• vault locations
• pruning or maintenance thresholds

What does not belong here:

• operator biography
• persona text
• life rules in prose
• Grok profile data
• freeform notes for retrieval

Why: CONFIG.json should be deterministic and operational. If a value changes how the software runs, it belongs here. If it changes how the agent understands the operator, it belongs in markdown identity docs instead.

TUI Update Contract

The aim tui "Update Operator Profile & Behavior" flow should remain symmetric with onboarding.

It should update:

• GEMINI.md for execution mode, cognitive level, conciseness, and lightweight guardrails
• core/OPERATOR.md for structured operator facts
• core/OPERATOR_PROFILE.md for narrative persona text

It should not write operator identity into:

• core/CONFIG.json
• memory/
• continuity/

Prompt-To-File Map

Behavioral prompts

Prompt: Grammar & Explanation Level

Destination: GEMINI.md / AGENTS.md

Reason: This controls agent presentation behavior, not runtime configuration.

Prompt: Enable Extreme Conciseness

Destination: GEMINI.md / AGENTS.md

Reason: This is prompt behavior.

Prompt: Execution Mode

Destination: GEMINI.md / AGENTS.md

Reason: This is agent operating posture.

Prompt: Target Model Intelligence

Destination: GEMINI.md / AGENTS.md

Reason: This controls prompt guardrails and instruction strictness.

Operator prompts

Prompt: Your Name

Destination:

• GEMINI.md / AGENTS.md
• core/OPERATOR.md
• core/MEMORY.md

Reason: The name is both prompt identity and durable operator identity.

Prompt: Core Tech Stack

Destination: core/OPERATOR.md

Reason: This is stable operator metadata.

Prompt: Working Style

Destination: core/OPERATOR.md

Reason: This is factual operator preference metadata.

Prompt: Metrics (Age/Height/Weight)

Destination: core/OPERATOR.md

Reason: This is structured operator profile data.

Prompt: Life Rules/Principles

Destination: core/OPERATOR.md

Reason: These are durable operator values.

Prompt: Primary Mission/Life Goal

Destination: core/OPERATOR.md

Reason: This is stable high-level operator intent.

Prompt: Business Info

Destination: core/OPERATOR.md

Reason: This is structured business context for the operator.

Prompt: Grok profile paste

Destination: core/OPERATOR_PROFILE.md

Reason: This is richer narrative persona material, not structured user metadata.

Environment prompts

Prompt: Obsidian Vault Path

Destination: core/CONFIG.json

Reason: This is runtime configuration.

Prompt: Allowed Root

Destination: core/CONFIG.json

Reason: This is runtime workspace policy.

Recommended Ingestion Semantics

• Root GEMINI.md or AGENTS.md: foundational operating prompt.
• core/*.md: foundational identity and durable project context.
• synapse/: external expert knowledge, reference packs, imported doctrine, and ingest-only material.
• memory/: refinement outputs, proposals, archival rollups, and evolving memory products.
• continuity/: checkpoints, summaries, and state-transfer artifacts.

Practical Distinction: OPERATOR.md vs OPERATOR_PROFILE.md

Use OPERATOR.md when the data should be:

• factual
• structured
• stable
• easy to diff
• easy to rewrite destructively

Use OPERATOR_PROFILE.md when the data is:

• interpretive
• narrative
• voice-oriented
• inferred from public behavior or social corpus
• helpful for stylistic alignment but not authoritative as fact

Short version:

• OPERATOR.md = canonical facts about the operator
• OPERATOR_PROFILE.md = richer persona interpretation of the operator
• CONFIG.json = runtime mechanics only