Back to Guides

A short AI-narrated film with English captions. Pressing play loads the YouTube player. Watch on YouTube

Guide Origin: Community

/arckit-agent-architecture:agent-maturity assesses an AI agent programme across design, governance, security, integration, and operations. It produces a current-state maturity view, target maturity, improvement roadmap, benchmarks, and traceability.


Inputs

Artefact Purpose
Agent inventory (ARC-<id>-AAGI-v1.0.md) Agent baseline, ownership, lifecycle, capability context
Agent design (ARC-<id>-AAGR-v1.0.md) Design patterns, tool contracts, guardrails, orchestration
Agent governance (ARC-<id>-AAOV-v1.0.md) Oversight, audit, monitoring, approval process
Agent security (ARC-<id>-AASE-v1.0.md) Security controls, threat model, validation pipeline
Agent integration (ARC-<id>-AAIN-v1.0.md) Integration patterns and interoperability evidence
Risk register (ARC-<id>-RISK-v1.0.md) Operational, compliance, and AI risk context
Enterprise maturity model Wider maturity baseline and benchmark context

Command

/arckit-agent-architecture:agent-maturity <project ID or agent programme>

Output: projects/<id>/ARC-<id>-AAMT-v1.0.md


Maturity Dimensions

Dimension Evidence
Design Reusable patterns, tool contracts, design reviews, ADRs
Governance Oversight model, audit coverage, approval records, compliance mapping
Security Threat model, sandboxing, permission controls, security testing
Integration Standard APIs, contracts, observability, failure isolation
Operations Monitoring, incident response, SLAs, support model, continuous improvement

Maturity Levels

Level Meaning
L1 Ad-hoc Reactive, informal, undocumented
L2 Reactive Some process after issues occur
L3 Defined Standard process documented and repeatable
L4 Managed Metrics-driven and actively governed
L5 Optimized Predictive, continuously improved, evidence-led

Deliverable Snapshot

Section Contents
Maturity Model Framework Dimensions, levels, and evidence criteria
Current State Assessment Level, evidence, gaps, and risk per dimension
Target State Target level and rationale per dimension
Improvement Roadmap Initiatives, owners, timeline, dependencies, success criteria
Benchmarks Cited industry, regulatory, and peer comparisons, each on the reference's own scale
Traceability Links to inventory, design, governance, security, integration, and roadmap artefacts

Review Checklist

  • Ratings cite evidence rather than relying on opinion.
  • Target levels are realistic for the investment and timeline.
  • Each improvement initiative has owner, success criteria, and dependency notes.
  • Security and governance maturity are not averaged away by strong design maturity.
  • Every benchmark cites its source, or says "No cited benchmark". NIST AI RMF and the OWASP Top 10 for LLM Applications define no maturity levels, so they are compared by Current and Target Profile and by risk coverage, not given a level.
  • Roadmap actions feed back into /arckit-agent-architecture:agent-design, /arckit-agent-architecture:agent-governance, and /arckit-agent-architecture:agent-security.

  • /arckit:maturity-model - Assess wider enterprise architecture maturity.
  • /arckit-agent-architecture:agent-design - Address design maturity gaps.
  • /arckit-agent-architecture:agent-governance - Address governance maturity gaps.