Commands Reference

AgentForge™ Commands Reference v1.0

Status: Stable (v1.0)
Scope: Shared control commands and conventions used across AgentForge agents.

1. Overview & Scope

The AgentForge™ Commands Reference defines how control commands are written, interpreted, and documented across all AgentForge-compliant agents.

Commands are plain-text control signals inside a conversation or interaction. They are not API calls but structured instructions that agents agree to recognize and obey.

1.1 Goals

Provide a consistent, predictable command language for all agents.
Separate control from content to reduce ambiguity.
Support diagnostics, math, export, and indexing across projects.
Allow individual agents to extend commands without breaking the core model.

1.2 Relationship to Standards v1.0

These commands align with the AgentForge Standards v1.0 page and assume projects follow the A–F folder model and agent structure described there.

2. Command Syntax & Conventions

2.1 Basic Syntax

#COMMAND_NAME [ARGUMENTS] [FLAGS]

Commands are written in UPPERCASE and prefixed with #.
A command should be on its own line whenever possible.
Arguments are optional and may be positional or key/value based.

2.2 Examples

#INIT
#DIAG FULL
#RUNMATH 3 * (7 + 5)
#EXPORT AGENT=SummarizerAgent FORMAT=markdown
#INDEX PROJECT

2.3 Interpretation Rules

If a line begins with # and matches a known command, the agent should treat it as a control instruction, not normal prose.
If an agent does not recognize a command, it should either:
- Ignore it and continue with normal behavior, or
- Report that the command is unknown (recommended for diagnostics agents).
Agents must not invent side effects beyond what is documented for that command.

3. Core Project-Level Commands

These commands are intended to be recognized by Root Agent and other project-wide agents.

3.1 #INIT

#INIT

Purpose: Initialize or reset project-level context.

Re-aligns the agent with the current project (e.g., AgentForge Framework).
Re-applies high-level standards, roles, and constraints.
Should be safe to run at any time.

3.2 #HELP

#HELP
#HELP COMMANDS

Purpose: Request a concise help summary.

#HELP — list core capabilities and key commands.
#HELP COMMANDS — show a focused list of commands and short descriptions.

3.3 #STANDARDS

#STANDARDS

Purpose: Ask the agent to respond in a way that explicitly respects AgentForge Standards v1.0.

Re-confirms A–F structure, naming, and lifecycle expectations.
Useful before generating project files, templates, or documentation.

3.4 #COMMANDS

#COMMANDS

Purpose: Ask the agent to list the commands it supports in the current context.

Should produce a concise, categorized list.
Diagnostics Agent is expected to provide a more detailed response than general-purpose agents.

4. Diagnostics & Health Commands

These commands focus on introspection, troubleshooting, and verification. They are primarily handled by the Diagnostics Agent but may be partially understood by others.

4.1 #DIAG

#DIAG
#DIAG FULL

Purpose: Run a diagnostic check of the agent’s understanding of the current context and expectations.

#DIAG — quick, high-level health check.
#DIAG FULL — more detailed, including assumptions, active standards, and key constraints.

4.2 #ECHOON / #ECHOOFF

#ECHOON
#ECHOOFF

Purpose: Control whether the agent explicitly echoes or restates certain inputs.

#ECHOON — agent may restate important instructions or context for clarity.
#ECHOOFF — agent should minimize repetition and keep responses tighter.

5. Deterministic Math & Utility Commands

These commands are handled by utility-style agents such as the RUNMATH Agent. Their purpose is to produce deterministic, stepwise outputs instead of approximate reasoning.

5.1 #RUNMATH

#RUNMATH <expression>

Purpose: Evaluate a mathematical expression step by step and return a deterministic result.

Agents should treat this as an instruction to calculate, not to guess.
Expected to show intermediate steps when helpful or requested.

Example

User:
#RUNMATH 275 * 16

Agent (RUNMATH mode):
1) 275 * 16
2) 275 * (8 * 2)
3) 275 * 8 = 2200
4) 2200 * 2 = 4400
Result: 4400

Exact formatting may vary by implementation, but the outcome must be deterministic and auditable.

6. Export & Index Commands

These commands are conceptually handled by utilities like CEUA (export/backup) and CIUA (index management). In conversational agents, they usually trigger structured output or instructions that a human or external tool can act on.

6.1 #EXPORT

#EXPORT
#EXPORT AGENT=RootAgent FORMAT=markdown

Purpose: Request a structured export of key artifacts.

May generate summaries, configuration snippets, or file stubs.
Parameters like AGENT= and FORMAT= are optional but recommended.

6.2 #INDEX

#INDEX
#INDEX PROJECT
#INDEX AGENT=SummarizerAgent

Purpose: Request an index view of known files, references, and dependencies.

#INDEX PROJECT — high-level project view.
#INDEX AGENT=Name — targeted index for a specific agent.

7. Content-Oriented Utility Commands

These commands support content transformation and are typically implemented by specialized agents such as the Summarizer Agent or documentation-oriented agents.

7.1 #SUMMARY

#SUMMARY

Purpose: Request a concise summary of the immediately preceding content or the current topic.

Summarizer Agent may offer additional parameters (e.g., length, style).

7.2 #OUTLINE

#OUTLINE

Purpose: Request an outline-style restructuring of the current topic or document.

7.3 #REFORMAT

#REFORMAT MARKDOWN
#REFORMAT HTML

Purpose: Request content to be restructured into a specific format.

#REFORMAT MARKDOWN — produce markdown-friendly output.
#REFORMAT HTML — produce HTML blocks suitable for web use.

8. Agent-Specific & Extended Commands

Individual agents may define additional commands beyond the core set above.

8.1 Definition Requirements

All agent-specific commands must be documented in that agent’s README.md and/or PROMPTS.md.
Recommended to list them under a section titled "Agent-Specific Commands".
Names should not conflict with the core commands unless behavior is intentionally extended.

8.2 Examples

#SCOUTMODE — a domain agent for Scouting scenarios might switch to a specific style.
#PREPPER_MODE — a preparedness agent could enable extra safety or redundancy checks.

These are examples only; each project defines its own domain-specific command set as needed.

9. Versioning & Compatibility

9.1 Command Set Version

The core commands documented on this page are Commands Reference v1.0.

Future versions (v1.1, v2.0, etc.) may add commands but will avoid breaking the meaning of existing ones.
Deprecated commands should be clearly marked and, where possible, mapped to replacements.

9.2 Per-Agent Compatibility

Not every agent is required to implement every command.
Agents should clearly declare:
- Which core commands they support.
- Any extended or agent-specific commands.

10. Summary

The AgentForge Commands Reference v1.0 defines a common control language for intelligent agents. By using clear, prefixed commands and documenting both shared and agent-specific behaviors, projects gain better governance, traceability, and predictability across their entire agent ecosystem.

AgentForge™ v1.0 Released

Today marks the completion of AgentForge™ v1.0 . This release establishes the foundational structure for building, documenting, and governing AI agents in a way that is consistent, reproducible, and maintainable over time. AgentForge is not a model, a chatbot, or a product layer. It is a framework —designed to bring order and clarity to how agents are defined, tested, promoted, and integrated. Version 1.0 deliberately focuses on fundamentals. It defines a standard project structure, a shared command language, and a starter set of templates that make expectations explicit from the beginning. Nothing in this release attempts to optimize creativity, speed, or novelty. Stability comes first. What v1.0 Includes AgentForge Standards v1.0 The authoritative definition of project structure, agent lifecycle states, governance rules, and documentation expectations. AgentForge Commands Reference v1.0 A shared command model that enables predictable control, diagnostics, and interaction p...

AgentForge™ – The AI Agent Framework