skills.rest
awslabs
awslabsOFFICIAL
Explore all skills by this author →
WebsiteGitHub

aidlc-portfolio

Coordinate parallel AI-DLC workflows across Git worktrees with a deterministic portfolio catalog.

1.2k|249|Updated Jul 29, 2025
Search Tags:#portfolio management#multi-agent orchestration#dependency graph#AI-DLC#Workflow Automation#Claude code#Git Worktrees
One-click install
npx skills add https://github.com/awslabs/cli-agent-orchestrator --skill aidlc-portfolio
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: aidlc-portfolio
Source: https://github.com/awslabs/cli-agent-orchestrator/tree/main/examples/aidlc-portfolio/skills/aidlc-portfolio
Command: npx skills add https://github.com/awslabs/cli-agent-orchestrator --skill aidlc-portfolio

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires ajv, yaml, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

Running multiple AI-DLC (AI-Driven Development Lifecycle) sessions across repositories and Git worktrees creates coordination chaos: untracked dependencies, unvalidated dispatches, lost human decisions, and conflicting shared-memory writes. This Skill provides a thin control plane that keeps portfolio state, child intents, and cross-project convergence deterministic and auditable.

Core Features & Use Cases

  • Portfolio Catalog & Lifecycle: Register projects, dependencies, and intents with evidence-backed confidence levels, then advance a persisted lifecycle (Bootstrap → Discover → Confirm → Plan → Dispatch → Integrate → Learn) gated by explicit human acceptance.
  • Deterministic Workspace Tooling: A Bun/TypeScript CLI (portfolio.ts) handles init, worktree creation, harness staging/sync/verify of Claude AI-DLC distributions, dispatch validation, and session tracking with locking and atomic writes.
  • Cross-Project Convergence & Human Relays: Child runners submit structured results; the tool computes dependency-graph impact, binds human decisions to risk revisions, relays AI-DLC questions to humans verbatim, and serializes shared-memory learning proposals under a portfolio lock.
  • Use Case: An operator names a repository and several work items; the supervisor bootstraps the workspace, creates one worktree per intent, projects a manifest-pinned Opus runtime into each, dispatches parallel aidlc_runner sessions via CAO, and synthesizes outcomes before integration.

Quick Start

Ask the agent to initialize an AI-DLC portfolio workspace for your repository and work items, then dispatch parallel child runners into validated Git worktrees.

Frequently Asked Questions about aidlc-portfolio

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run multiple AI-DLC workflows in parallel across repositories?▼

Initialize a portfolio workspace, register each project and intent, create one Git worktree per intent, then dispatch one aidlc_runner session per worktree using CAO assign. The portfolio.ts utility validates dispatch readiness and tracks each session's terminal ID.

How do I manage Git worktrees for parallel AI coding agents?▼

Use the worktree create command, which validates the registered repository and branch mapping before running git worktree add. Each child intent gets an isolated path under the portfolio root, and dispatch check confirms harness verification before assignment.

Does this work with Claude Code and the CAO orchestrator?▼

Yes. The harness commands stage a Claude AI-DLC distribution, pin the Opus model overlay, and project it into every worktree. Dispatch uses CAO assign with the aidlc_runner profile and an absolute working_directory per worktree.

How are human decisions handled during automated AI-DLC stages?▼

Child runners submit unanswered question packets and stop; the supervisor collects verbatim human answers and resolves the packet with recorded provenance. Discovery confirmation, plan acceptance, and convergence risk decisions all require named human approvers bound to catalog revisions.

What happens if a runner terminal crashes mid-session?▼

CAO terminals are treated as disposable while AI-DLC state is durable. Run the status command, reconcile against the registered worktree and child state, then relaunch the same runner against the same intent without creating replacement workflow state.

Why does session completion fail with shared-memory errors?▼

Completion is blocked while project.md or team.md changes remain in the feature worktree. Submit learning proposals for supervisor approval, then run memory inspect and memory clean with the exact worktree revision before marking the session completed.

Related Skills

Explore More Software Engineering Skills→
stephanesoaresstephanesoares

fix-github-issue

Automate GitHub issue fixes by creating feature branches, running tests, and opening pull requests.

Community
Intermediate
robpitcherrobpitcher

git-workflow

Define a three-branch Git workflow with main, dev, and insiders branches.

Community
Intermediate
dx-junkyarddx-junkyard
1

episteme-graph-ci-tests

Generate pytest unit tests for Episteme Graph API routes, schemas, and agent logic.

Official
Intermediate
amadeus-dlcamadeus-dlc
5

amadeus-scope-definition

Run the AI-DLC scope-definition stage as an isolated single-stage workflow.

Official
Intermediate