Skip to main content

Components

7 types | 327 visible | Post-5S: KISS/DRY/LEAN

The ADLC framework organizes capability into 7 component types. Each serves a distinct purpose in the agent governance lifecycle.

At a Glance​

TypeIconCountPurposeWho Uses It
Agents🤖38AI team members with defined roles and authorityHITL manager delegates to agents
Commands⚡117Slash commands that automate multi-step workflowsDevelopers and agents invoke /command
Skills⚙️94Reusable capability packages with bundled referencesAgents load skills for domain expertise
Hooks🪝12Governance scripts that enforce rules before/after actionsSystem auto-triggers on tool calls
MCP Servers🔌58External API integrations via Model Context ProtocolAgents call MCP tools for cloud/data access
Plugins🧩8Pattern libraries and reference architecturesAgents reference patterns during implementation
Settings🔧60Claude Code configuration presetsFramework bootstrap applies settings

Agents​

AI team members with defined roles, authority boundaries, and tool access. Enterprise governance requires separation of concerns — each agent has a specific accountability scope. The framework includes 9 core constitutional agents and 30 marketplace specialists across 11 categories.

See Agents Overview for details.

Commands​

Slash commands (/command-name) that automate multi-step developer workflows. 74 core commands span 14 domains (terraform, cdk, finops, k3d, k3s, docs, security, and more), plus 51 marketplace commands across 7 categories including product management.

Skills​

Reusable capability packages with bundled reference docs, scripts, and templates. Skills give agents deep domain expertise without loading everything into context. 20 core skills across 6 domains (development, dashboards, cdk, testing, finops, governance) plus 75 marketplace skills across 6 categories including 65 product management skills.

Hooks​

Governance enforcement scripts that auto-execute before and after agent tool calls. Hooks are the hard guardrails — they block actions that violate constitutional principles. 12 scripts enforce coordination, delegation, security, and evidence requirements across 6 layers (prompt, tool, post-execution, session, permissions, rules).

See Hook Enforcement Reference for the complete coverage matrix, enforcement chain details, and file-lock mechanism.

MCP Servers​

External API integrations via the Model Context Protocol standard. Each server exposes tools that agents can call for real-time data from cloud providers, databases, and services. 24 enterprise servers configured with 58 marketplace definitions available.

Plugins​

Pattern libraries and reference architectures providing domain knowledge. 8 plugins cover database patterns, deployment operations, security, testing, context optimization, drift detection, and experiment tracking.

Settings​

Claude Code configuration presets organized by concern. 60 JSON configuration files across 11 categories ensure consistent guardrails, model preferences, and tool permissions across enterprise teams.

Component Lifecycle​

PLAN          BUILD              TEST              DEPLOY
  │              │                  │                  │
  ▼              ▼                  ▼                  ▼
Agents ──→ Commands + Skills ──→ Hooks enforce ──→ MCP validates
(coordinate)  (execute)         (governance)       (cross-validate)
                  │
                  ▼
            Plugins + Settings
            (reference patterns)

Flow: HITL delegates to agents → agents invoke commands → commands use skills for domain knowledge → hooks enforce governance at every tool call → MCP servers provide real-time data → plugins supply reference patterns → settings ensure consistent configuration.

Deep-dive references: