Skip to content

Latest commit

 

History

History
25 lines (17 loc) · 2.16 KB

File metadata and controls

25 lines (17 loc) · 2.16 KB

Agent Registry

ARCHON ships a single agent. This registry exists as the canonical record of that agent's identity and scope, per ADR-001 in ARCHITECTURE_DECISIONS.md (single cohesive agent with internal L1-L5 levels, rejecting a separate-agent-per-C-suite-role model).

Field Value
Name archon
File agents/archon.md
Persona Principal Engineer / CTO-level technical decision-maker
Levels L1 Foundations → L2 Software Engineering → L3 Infrastructure & Cloud → L4 Principal Engineering → L5 CTO & Business
Tools Read, Grep, Glob, WebSearch, mcp__github__* (optional, read-only — see ADR-005)
Model inherit
Invocation Via the 9 slash commands in COMMAND_REGISTRY.md, or directly as a subagent

Why One Agent, Not Many

The source playbook this project draws structural inspiration from describes a full C-suite roster (CEO/CPO/CTO/CIO/COO/CFO/CRO/CMO/CHRO/CBO agents). ARCHON deliberately implements only the CTO/Principal-Engineer persona as a single agent with internal seniority levels, because the actual need this project serves is a single, consistent engineering decision-making voice — not a simulated executive team. See ADR-001 for the full reasoning.

How the Agent Uses the Skill Library

The archon agent does not hardcode domain knowledge in its own definition. It reasons using the Engineering Decision Principles and the 8 Core Principles defined in skills/00_Core/, and pulls domain-specific decision rules from the 20 domain skill modules (skills/01_Foundations/ through skills/19_Review_Outputs/) plus the cross-cutting skills/99_Decision_Engine/ routing layer, exactly as a Claude Code subagent is expected to compose with the skill library available to it. See SKILL_REGISTRY.md for the full skill module list and MODULE_INDEX.md for the complete repository map.

Adding a New Agent

Per CONTRIBUTING.md, new agents are out of scope unless they represent a genuinely distinct persona need (not just a new domain of knowledge, which belongs in skills/ instead). Any proposal to add a second agent should start with an ADR explaining why the single-agent model in ADR-001 no longer fits.