# A.D.A.M.

Name: A.D.A.M.
Subtitle: Adaptive Depth and Mode
Type: AI control protocol
URL: https://xxyoudeadpunkxx.github.io/A.D.A.M.-Adaptive-Depth-and-Mode/
Repository: https://github.com/XxYouDeaDPunKxX/A.D.A.M.-Adaptive-Depth-and-Mode
License: CC BY-SA 4.0
License URL: https://creativecommons.org/licenses/by-sa/4.0/
Canonical file: A.D.A.M.v5.txt

## What this is

A.D.A.M. is a spec-first protocol for controlling response depth, preserving signal, and reducing noise in AI chat.

It is not a chatbot personality, prompt aesthetic, or generic productivity framework. It is a control layer for routing answers through explicit depth modes and bounded audit behavior.

## Operating line

Simple work stays simple. Complex work can escalate. Ambiguous work is routed before it drifts.

## Use case

Use A.D.A.M. when AI chat responses become too shallow, too deep too early, too assumptive, or too noisy for the actual task.

The protocol is designed for everyday chat, structured tasks, and deeper analysis when needed.

## Canonical source

The canonical protocol file is:
A.D.A.M.v5.txt

Treat the repository as the live public source. Treat this file as the LLM-readable project index, not as the full protocol specification.

## Core behavior

LOW:
Fast, minimal answers.

MID:
Normal structured answers.

POSSIBLE DEEP:
A deeper pass may be useful. The protocol proposes escalation instead of silently expanding.

DEEP:
Full decision support through automatic detection plus consent, or through manual override.

Bounded audit:
When audit is active, ACTION, RISK, and BASIS remain short and operational.

Fail-closed behavior:
Unsupported host conditions can produce ADAM_UNSUPPORTED instead of pretending the protocol is active.

## Activation summary

Load A.D.A.M.v5.txt into a compatible chat host instruction surface.

At session start, send ADAM PING twice.

The first ping starts the protocol and checks bootstrap state.
The second ping confirms liveness on the next turn.

## Commands

- ADAM PING: activate or check liveness.
- LOW: force a short answer.
- MID: force a normal structured answer.
- DEEP: force full decision support.
- ADAM SELF TEST: run the external self-test if available.
- SYS STATUS: run the status probe if available.
- TRACE INPUT: show a strict trace of the last visible user input.
- ADAM OFF: turn the protocol off.
- ADAM ON: turn the protocol back on.
- ADAM REMOUNT: generate a recovery capsule for remounting.

## Repository files

- A.D.A.M.v5.txt: canonical protocol file.
- ADAM_SELF_TEST.md: external self-test probe.
- SYS_STATUS.md: lightweight session health/status probe.
- docs/lexicon.md: official terminology.
- docs/field-report-chat.md: field report from a real creative workflow.

## Discovery Set

Machine-readable project files:

- llms.txt: LLM-readable project index.
- raw-manifest.json: structured machine-readable project manifest.
- robots.txt: crawler access declaration.
- sitemap.xml: sitemap for the GitHub Pages surface.

HTML discovery signals:

- canonical URL
- rel=alternate for llms.txt
- rel=alternate for raw-manifest.json
- JSON-LD SoftwareSourceCode metadata
- Open Graph metadata

## Keywords

AI protocol, AI chat control, response depth, LOW MID DEEP, bounded audit, prompt protocol, A.D.A.M., Adaptive Depth and Mode, AI response routing, ambiguity control, structured AI chat