AITF Protocol v4.0

Purpose

AITF is the universal expression language for agent-to-agent communication. An agent loads this protocol, and it can read any AITF document and write any AITF document. The same rules govern both input and output — no separate "read format" and "write format".

1. File Rules

One File = One Complete Context

All structured content lives in .md files. Any agent reading a file can understand the full state without external information. No chatroom, no message queue, no out-of-file discussion.

Record Format

Each record is one line. Fields are separated by pipe |. Hierarchy is expressed by explicit parent reference ^parent_id.

format: id|type|status|content|^parent_id|depends|impacts|author

example: A1.2|A|U|user_retention_driven_by_onboarding_quality|^C1|D1.1|S2,K3|@McMillan

File Header (Mandatory)

Every AITF document must begin with a META block:

## META
protocol_version: v4.0
type: {proposal|knowledge|task|suggestion|hybrid|other}
scope: {what this document covers}
status: draft|in_progress|review|closed
author: @{AgentName}
participants: [{agent list if multiple}]
created_date: {YYYY-MM-DD}
source: {optional: origin material if derived from external}
source_author: {optional: original author}

2. Node Type Registry

All node types are available for any document. The agent selects which types to use based on what it needs to express. Node types are not locked to scenarios — a task document can include risk nodes, a knowledge document can include gap nodes, a proposal can reference extracted concepts.

Analysis Nodes

Used to express reasoning, arguments, and solutions.

Knowledge Nodes

Used to express structured knowledge extracted from sources.

Task Nodes

Used to express work assignments and execution records.

Collaboration Nodes

Used to express suggestions, responses, and inter-agent dialogue.

Mandatory Node Attributes

Every node must carry these four attributes:

3. Status Codes

Universal Status Codes

These apply to all node types across all documents.

Status Assignment Rules

Status Transitions

U → V :  independent verification provided
U → X :  conflicting evidence or interpretation found
U → F :  agent cannot resolve, needs human
N → U :  partial basis found
N → V :  full evidence found and verified
X → V :  conflict resolved with sufficient evidence
V → X :  new contradicting evidence (must be substantiated)
any → W :  author retracts (must state reason)
B → V :  dependency resolved
B → F :  dependency unresolvable by agents

4. Structure Rules

Hierarchy

Nodes form trees via parent reference. Depth is unlimited — the agent uses whatever depth the content requires.

P1[U]: problem statement
  E1.1[N]: evidence
  A1.1[U]: assumption
  C1.1[U]: claim
    S1.1.1[U]: solution
      M1.1.1.1[U]: mechanism
      K1.1.1.1[U]: risk

References

Nodes can reference other nodes for traceability. All references must point to existing nodes — no orphan references.

Relation Types (for REL nodes)

Tag Rules (for TAG nodes)

Each TAG must include synonyms and aliases, likely query phrasings (how an agent would search for this), and domain-specific terms if applicable.

Good: covers ≥3 search angles (formal term, colloquial, abbreviation, related domain)
Bad:  single word or only exact name

Example good: [marginal cost, MC, incremental cost, variable cost per unit, how much does one more cost]
Example bad:  [marginal cost]

5. Signing and Versioning

Authorship

Every piece of content must be signed: @{AgentName}. When multiple agents contribute to one document, each agent's contribution is clearly attributed. No agent modifies another agent's signed content. To disagree: write your own signed node (not edit theirs).

Document Versioning

Changelog (Mandatory)

Every document ends with a changelog:

## CHANGELOG
v1.0 @AgentName: {what was done}
v1.1 @AnotherAgent: {what was added}

6. Quality Standards

Self-Check (Before Finalising)

The author must verify before finalising any document:

• Every node has id, type, status, author (no missing attributes)
• All references point to existing nodes (no orphans)
• Status assignment rules followed (no self-verified A/C/S/M/K)
• No circular dependency chains (A based on C based on A)
• Document is self-contained (reader needs no external info)

Anti-Patterns

Honest Assessment (Recommended)

For substantive documents, the author appends a self-assessment:

## SELF_ASSESSMENT
weakest_node: {ID}: {why this is the weakest point}
most_uncertain_assumption: {A_ID}: {why}
biggest_gap: {G_ID}: {why}
overall_confidence: high|medium|low

Gap Severity

G_ID[open]: {description}
  severity: critical|high|medium|low
  blocking: [{node IDs that cannot be V without this}]
  resolution_hint: {what might fill this gap}

7. Escalation

When to Mark F (Flag for Human)

• Agents contradict with valid evidence and no resolution path
• Requires empirical test or runtime validation
• Commercial judgement (cost, priority, resource allocation)
• Irreversible action (cannot undo after execution)
• Ambiguous source material (multiple valid interpretations)
• Concept boundary unclear (needs domain expertise)
• Cross-source contradiction (same topic, different claims)
• Scope or priority judgement (what to include or exclude)
• Touches agent identity or SOUL.md

Escalation Format

node_ID[F]: {description}
  @{AgentName}: escalation
    reason: {why agents cannot resolve}
    options: [{possible resolutions}]
    human_question: {what specific decision is needed}

8. Multi-Agent Collaboration

Working on the Same Document

• Each agent reads the full existing document first
• Each agent writes new nodes signed with own @{AgentName}
• No agent modifies another agent's content
• To disagree → write own node with counter-argument or alternative
• Disagreement unresolvable → mark F

Cross-Document References

When referencing nodes in another document, use the format:

{document_filename}:{node_ID}

example: project_alpha_proposal.md:S2.1

9. Backward Compatibility

Old files must remain readable under new protocol versions.

10. Protocol Governance

Modification Process

• Agents propose changes via a protocol opinion document
• Human reviews and approves
• Agents NEVER modify the live protocol file directly

Feedback During Work

When an agent discovers a protocol gap during a task:

1. Complete the current task normally (do not interrupt)
2. Record the gap in a protocol opinion document with severity rating
3. Continue working

G_ID[open]: {gap description}
  trigger: {task that exposed this}
  date: {YYYY-MM-DD}
  severity: critical|high|medium|low

Links

• GitHub — AITF-Agent-Interaction-Text-Format
• IAF Framework — Intelligenism Agent Framework
• Website — intelligenism.club
• Theory — intelligenism.org