Skip to content

Agents — Taxonomy, Communication, and Identity

The section that addresses "who, what, on whose behalf, and how they coordinate." Deepens the Agent layer of the three-layer model (03-architecture).

Position of This Section

The Agents section is being grown incrementally toward the Agent ID era. It is organized: terminology → implementation units → communication protocols → identity.

Page Overview

Concept layer — Get terminology straight first

PageThemeBest for
Agent TaxonomyOrganize custom/sub/meta-agents, Orchestrator-Worker, Swarm, etc., into 4 abstraction levelsAnyone wanting cross-framework terminology

Implementation layer — What to build and how

PageThemeBest for
Custom Sub-agentSub-agent definition, placement, and use in Claude CodeFoundational understanding before implementation
Sub-agent vs SkillsWhether to implement as a Skill or a Sub-agent; composition patternsWhen you're stuck on a design decision
Quality Gate PatternValidator sub-agent implementation, CI/CD integration, pass criteriaWhen self-review keeps coming out too lenient
Multi-Agent / Agent TeamsBeyond single sub-agents — Orchestrator-Worker, Hierarchical, Swarm patternsWhen single agents hit a ceiling

Communication layer — How agents talk to each other

PageThemeBest for
What is A2A (Agent-to-Agent Protocol)A2A v1.0 overview, standardization under Linux Foundation, complementarity with MCPAnyone exploring cross-org agent coordination

Identity layer — On whose behalf do they act?

PageThemeBest for
Agent IdentityNon-Human Identity as a new category, OpenID Foundation's 4 architectural philosophies, commercial implementation statusOperators bringing agents to production

Roadmap Toward the Agent ID Era

This site grows alongside the 2025–2026 Agent ID standardization wave. Here is the honest current status and what's coming.

TopicStatus
Agent Taxonomy✅ Published
Custom Sub-agent (Claude Code)✅ Published
Sub-agent vs Skills✅ Published
Quality Gate Pattern✅ Published
Multi-Agent / Agent Teams✅ Published
A2A Protocol✅ Published
Agent Identity (identification & delegation)✅ Published
Permission management (RBAC / ABAC / JIT)🚧 Planned — referenced from agent-identity
Agent governance🚧 Planned
A2A implementation patterns (Web Bot Auth, Macaroons, etc.)🚧 Planned
Agent Marketplace / Registry🚧 Under consideration

IMPORTANT

The Agent ID area has rapidly moved into production-operation phase: OpenID Foundation white paper v1.1 (October 2025), Microsoft Entra Agent ID GA (April 2026), Okta for AI Agents GA, A2A v1.0 under Linux Foundation. Because specs remain fluid, this site distinguishes "what is settled now" from "what remains open."

Route 1: Newcomer (foundation)

Route 2: Production readiness (identity-first)

Route 3: At scale (multi-agent coordination)

Route 4: Raising quality (implementation techniques)

SectionRelation
Concepts / 03-architectureThe whole three-layer model including the Agent layer
Concepts / 07-doctrine-and-intentDesigning "constraints and objectives" given to agents
Concepts / 08-memory-and-knowledgeThe memory layer agents reference
SkillsStatic knowledge agents reference
MCPHow agents connect to the outside world
FAQ / Agent vs Sub-agent vs Skill vs MCPThe 3-line answer for all four

🔗 Deeper: From the context-management perspective

For agents' essential constraints — why isolated contexts matter, why Multi-Session coordination is needed — grounded in LLM structure, see the sister site.

Released under the MIT License.