Tried running the newly GA'd AI-DLC Workflows v2.5.5 with Kiro CLI
This page has been translated by machine translation. View original
Introduction
We published a preview article for AI-DLC Workflows v2.0.0 in June 2026. About five weeks later, v2.0 reached GA.
The current version v2.5.5 is the latest release, incorporating minor updates since GA.
In this article, we use dist/kiro/ from v2.5.5 to launch a workflow with Kiro CLI and verify operation up to the question generation and answer mode presentation in the intent-capture stage.
Verification Details
Verification Environment
| Item | Value |
|---|---|
| Kiro CLI | 2.13.1 |
| AI-DLC Workflows | v2.5.5 |
| Conductor Model | claude-opus-4.6 |
| Sub-agent Model | claude-sonnet-4.6 |
| bun | Kiro CLI built-in version (used by adding to PATH) |
| Launch Method | Using pre-built dist/kiro/ |
The AI-DLC defaults are claude-opus-4.8 for the conductor and claude-sonnet-4.5 for the sub-agent. In the author's environment, the defaultModel in cli.json and each agent definition were modified to evaluate with the opus-4.6 / sonnet-4.6 combination.
Since dist/ is committed as pre-built, it can be placed simply by cloning and copying.
git clone -b v2 https://github.com/awslabs/aidlc-workflows.git /tmp/aidlc-v2
cp -r /tmp/aidlc-v2/dist/kiro/.kiro <project>/.kiro
cp -r /tmp/aidlc-v2/dist/kiro/aidlc <project>/aidlc
cp /tmp/aidlc-v2/dist/kiro/AGENTS.md <project>/AGENTS.md
Major Changes from v2.0.0 → v2.5.5
The structural changes from v2.0.0 to v2.5.5 are as follows.
| Item | v2.0.0 | v2.5.5 |
|---|---|---|
| Number of Stages | 32 | 32 |
| Number of Agents | 13 | 14 (+composer) |
| Number of Harnesses | 4 | 5 (+opencode) |
| Audit events | 67 | 74 |
| Number of Scopes | 9 | 9 |
| Stage modes | 2 (inline, subagent) | 4 (+pipeline, +mob) |
Key new features are listed below.
- Adaptive Workflows (v2.2.0) The composer agent can now infer scope from natural language input and dynamically generate EXECUTE/SKIP grids
- Plugin Mechanism (v2.3.0) A mechanism for defining extended stages and contributions in the
plugins/directory was added - opencode Harness (v2.4.6) Support for opencode.ai was added as the fifth harness
- 3-Role Ensemble (v2.5.0) Collaborator separation via
pipeline/mobmodes and contribution files were introduced
Summary of Changes by Major Version
| Version | Date | Overview |
|---|---|---|
| 2.1.0 | 06-24 | Full workspace overhaul: aidlc-docs/ → aidlc/spaces/default/intents/. Per-intent layout, spaces/intents/multi-repo |
| 2.2.0 | 07-04 | Adaptive Workflows: composer agent (14th). Automatic scope selection via NL keyword inference, dynamic EXECUTE/SKIP grid generation |
| 2.3.0 | 07-07 | Plugin mechanism: extended stage and contribution definitions via plugins/ |
| 2.5.0 | 07-17 | 3-role ensemble: pipeline/mob modes. Collaborator separation, contribution files |
Running doctor
After placement, running /aidlc --doctor resulted in 39 passed, 0 failed.
Full doctor Output
AI-DLC Health Check
─────────────────────────────────────
✓ bun installed (required for CLI tools and hooks)
✓ aidlc-audit-logger.ts present
✓ aidlc-sync-statusline.ts present
✓ aidlc-validate-state.ts present
✓ aidlc-log-subagent.ts present
✓ aidlc-session-start.ts present
✓ aidlc-session-end.ts present
✓ aidlc-statusline.ts present
✓ aidlc-kiro-adapter.ts present
✓ agents/aidlc.json present (hook + permission wiring)
✓ settings/cli.json present (workspace default-agent activation)
✓ AWS_AIDLC_DEFAULT_SCOPE (unset — no project default)
✓ Enabled plugins: all enabled (no selection); enabled stage counts: aidlc=29, bootstrap=3
✓ Plugin selection flags: harness.json agrees with stage-graph.json
✓ Enabled stage compile coverage: every enabled stage file is in the full graph
✓ workspace shell ready (.kiro/ + aidlc/spaces/default/memory/)
✓ Agent filename/name consistency: all agent files match declared names
✓ Scope filename/name consistency: all scope files match declared names
✓ Submodules: no .gitmodules at workspace root
✓ Hooks last fired: ...
✓ Hook drops: none recorded
✓ Audit locks: none leaked
✓ State Version: 7
✓ Orphan worktrees: 0 observed
✓ Stale branches: 0 (0 bolt-* observed)
✓ Orphan state files: 0 observed
✓ Orphan audit: 0 observed
✓ Practices staleness: never affirmed (informational)
✓ MERGE_DISPATCH: 0 orphan INVOKED (0 bracketed)
✓ Cycle detection: 0 cycles
✓ Orphan stage files: 32 graph entries all have files
✓ Uncompiled stage files: 0 stage files missing from the compiled graph
✓ Scope validation: 9 scopes valid (27 advisories)
✓ Schema validation: 32/32 stages validated
✓ Graph references: 122 artifacts + edges resolved
✓ Keyword overlap: no conflicts
✓ Rule drift: no team/project rule overlaps org policy
✓ Paired sensor coverage: no sensor-bound rules (0 feedforward-only)
✓ Intent registry: all rows ⇄ record dirs reconciled
─────────────────────────────────────
39 passed, 0 failed
Launching the Workflow
We launched the workflow using a calculator app as the subject.
/aidlc "シンプルな電卓Webアプリを作りたい"
The AI-DLC workflow engine presented scope options.
How would you like to proceed?
- compose — Compose a tailored workflow (adaptive)
- mvp — Skip operations, ship the core
- feature — New feature, practical depth
- poc — Proof of concept (8/32 stages)
- (other scopes available)
To verify the entry point with a minimal configuration, we selected poc (8/32 stages). intent-birth was executed and the workflow record was created.
intent-birth --scope poc --arguments "シンプルな電卓Webアプリを作りたい" --label "simple-calc"
Record created: aidlc/spaces/default/intents/260722-simple-calc/
The intent-capture stage then began, and the question file was generated.
Generated: intent-capture-questions.md (3 questions, PoC scope minimal depth)
Q1: Calculator purpose (for learning or practical use)
Q2: Supported operations (basic arithmetic only or scientific calculator)
Q3: Target users (developer themselves or end users)
Select answer mode:
- Guide me (interactive Q&A)
- I'll edit the file (direct edit)
- Chat (freeform conversation)
Verification was stopped at this point. We confirmed that the intent-capture stage operates correctly up to question generation and answer mode presentation.
Summary
AI-DLC Workflows v2.5.5 worked immediately on Kiro CLI with just a copy from dist/. The v2 architecture, Harness Engineering design philosophy, and usage are explained in detail in the article below, so we recommend reading it before evaluation.
