aidlc-deploy

name	aidlc-deploy
description	CI/CD pipeline generation and deployment configuration. Produces pipeline configs, environment definitions, and deployment scripts for the target platform.
license	MIT
compatibility	Requires file system access. Auto-detects environment.
metadata	{"author":"AI-DLC Maintainers","keywords":"specification, deploy, CI/CD, pipeline, infrastructure, environment, AI-DLC","supported_platforms":["kiro-ide","kiro-cli","claude-code"]}

Deploy Skill

Base: shared/base.md (full on first load, §Summary on chain). Actions: load per-step from actions/.

You generate deployment configurations and CI/CD pipelines. Translate the verified build into a deployable artifact with proper environment promotion, secrets management, and rollback strategy. You produce config files — you don't execute deployments directly.

When active:

1. Follow ONLY the process below
2. WAIT for user approval at each checkpoint
3. Never narrate your internal process
4. ALL output in the user's language (read manifest language field) — no English narration

---

Activation

✅ aidlc-deploy active — {platform} detected.
Ready to generate deployment configuration and CI/CD pipeline.

---

Quick Start

1. Generate D5 decision gate (deployment strategy, target platform, environments)
2. Validate decisions for conflicts → resolve if any
3. Generate deployment specification (pipeline architecture, infrastructure topology, promotion flow)
4. Present spec → wait for approval
5. Generate CI/CD pipeline configuration, environment definitions, and deployment scripts
6. Present results → wait for approval → mark workflow complete

Reads: build-report.md, design/implementation.md, context.md, blueprints, existing CI configs Writes: decisions-deploy.md, deployment.md (specs), pipeline config files, environment configs

---

Information Contract

Required Inputs

Information	Description	Accepted Formats
Build report	Verified build status and test results	Markdown (build-report.md)
Project context	Stack, architecture, ecosystem	Markdown (context.md), manifest
Design documents	Infrastructure and deployment needs	Markdown (design/implementation.md)

Optional Inputs

Information	Description	Accepted Formats
Existing CI config	Current pipeline to extend or replace	YAML (.github/workflows/*.yml, .gitlab-ci.yml, etc.)
Infrastructure code	Existing IaC to reference	Terraform, CDK, Pulumi, CloudFormation
Environment variables	Required env vars from implementation	.env.example, config files
Operations design	Health probes, logging config, metrics, shutdown settings	Markdown (design/operations.md)

Outputs

Artifact	Default Path	Description
decisions-deploy.md	{WORKFLOW_DIR}/{feature}/decisions-deploy.md	D5 decision gate answers
deployment.md	{SPECS_DIR}/{feature}/deployment.md	Deployment specification (pipeline, infra, promotion, risks)
Pipeline config	Project root (platform-specific)	CI/CD pipeline definition
Dockerfile	Project root	Multi-stage container build (if container target)
Environment configs	.env.{env}.example	Per-environment variable templates
Infrastructure code	infra/	Terraform or CDK files (if IaC selected in D5)
Deploy/rollback scripts	scripts/	Manual deploy and rollback helpers
deploy-summary.md	{WORKFLOW_DIR}/{feature}/deploy-summary.md	Deployment configuration summary

---

Initialization

1. Detect environment (per shared base)
2. Resolve feature name (per shared base)
3. Read manifest at {WORKFLOW_DIR}/{feature}/aidlc-manifest.yaml
4. Read {WORKFLOW_DIR}/{feature}/build-report.md for build context
5. Check for existing CI/CD configs (see detection list in references)
6. Read design/implementation.md for infrastructure requirements
7. Read design/operations.md if it exists (for health probes, logging, metrics, shutdown config)
8. Read blueprints Summary sections

---

Process

Execute actions sequentially. Load the action file when you reach that step — not before.

Step	Action	Load
1	Generate D5 decisions + validate	{SKILL_DIR}/actions/decision-gate.md
2	Generate deployment specification	{SKILL_DIR}/actions/plan.md
3	Generate pipeline and configs	{SKILL_DIR}/actions/generate.md
4	Finalize deployment	{SKILL_DIR}/actions/finalize.md

---

References

Load conditionally based on D5 answers — never load all references at once:

Reference	Load When
platforms.md	Always (overview/fallback for unsupported platforms)
platform-github-actions.md	D5 CI = GitHub Actions (read stack + target section only)
platform-gitlab-ci.md	D5 CI = GitLab CI (read stack + target section only)
iac-terraform.md	D5 IaC = Terraform (read target + database section only)
iac-cdk.md	D5 IaC = CDK (read target + database section only)
strategies.md	During decision gate (deployment strategy context)

---

Skill Handoff

No next skill — this is the final phase. On completion, mark workflow as complete.

---

Phase-Specific Rules

• This is a project-wide phase — always runs at the feature level, never per-unit.
• Audit actions: decision-gate, plan, validation, generation, approval, deploy-complete.

Deployment Rules

• Generate config files — never execute actual deployments
• Respect existing CI/CD configs: extend or replace based on user decision
• Always include rollback strategy in generated configs
• Secrets must use platform-appropriate secret management (never hardcoded)
• Include health checks and smoke tests in deployment pipeline
• Environment promotion order: dev → staging → production (configurable)

Security

• Never write secrets, tokens, or credentials into generated files
• Use placeholder references: ${{ secrets.NAME }}, ${SECRET_NAME}, etc.
• Flag any secrets that need to be configured in the CI/CD platform

---

Context Recovery

If context is lost mid-phase, follow aidlc/shared/base.md Context Recovery, then:

• Check if decisions-deploy.md exists at {WORKFLOW_DIR}/{feature}/:
  • Not present → load actions/decision-gate.md (start from D5)
  • Exists but no deployment.md at {SPECS_DIR}/{feature}/ → load actions/plan.md
  • deployment.md exists but no pipeline config generated → load actions/generate.md
  • Pipeline config exists but manifest artifacts.deploy.status is NOT "approved" → load actions/generate.md, re-present the generated files for approval (step 8). Never route to finalize.md here — finalize runs only after the user approves.
  • Manifest artifacts.deploy.status: "approved" → finalize already ran (it sets this together with state.status: "completed"); present workflow-complete status