Project Delivery Engine — Continuity for Long-Running AI Projects

Keep the project, not the chat, in control

Long-running AI work often loses momentum when a session ends, a new agent takes over, or several people and agents work in parallel. Project Delivery Engine keeps the durable project record in ordinary files: what is true now, what happens next, what is at risk, what evidence supports a decision, and where the next session should resume.

It is a downloadable Agent Skill built around SKILL.md, supporting references, and an included Node.js helper. It is not a standalone application or hosted service.

Before you buy

Your AI-agent host must be able to install a local SKILL.md Skill, read and write files in your project, and run Node.js 24 LTS or a later supported LTS release. Multi-agent delegation, independent review contexts, and Git worktrees are available only when your host provides those features.

For a standard long-running project, the workflow starts with five small state files and three indexes. It adds other records only when the project actually needs them.

What using it looks like

Start a project and let the Skill record the current goal, next action, risks, and handoff point. Days later, ask to continue and the next session resumes from those files. When the work grows, delegate bounded tasks and collect evidence with the results. Review important work independently, then prepare a clean handoff before moving to another session or agent.

Preserve knowledge that will matter again

Version 0.5.0 adds a practical project-knowledge workflow. It can retain verified deployment steps, agreed project terminology, a rejected approach and why it failed, or another lesson that future work would otherwise have to rediscover.

It keeps that knowledge in the smallest useful place, points back to one reliable source, and does not turn every conversation into a knowledge base. New sessions load only what matters to the task at hand.

Delegate work without losing control

• Define delegated work with a clear scope, acceptance criteria, artifact list, and stopping condition.
• Return evidence with the result so the coordinating agent can review what actually happened.
• Use an independent context to challenge important work before adopting it.
• Keep current status separate from history, evidence, materials, and handoff notes.
• Resume after a pause, a new session, or an agent change without rebuilding the project from old conversations.

Project Delivery Engine does not run agents by itself.

What is included

• The installable Project Delivery Engine Agent Skill.
• References for project state, knowledge, delegation, evidence, review, and handoff workflows.
• The project-gov helper for inspection, validation, migration checks, material checks, identifiers, startup prompts, and controlled record updates.
• Installation and operating documentation.
• SHA-256 checksums for the buyer package.
• A single-user commercial license.

Compatibility

The core instructions and project files are ordinary text. Installation paths and Skill discovery vary by host; the buyer guide explains the installation and verification steps.

Price and license

USD 50 one-time purchase.

One purchase grants a perpetual license to one individual for unlimited personal or commercial projects on devices they control. Local modifications are allowed for that buyer's own use.

The package and modified copies may not be shared, sold, sublicensed, publicly mirrored, or redistributed. Project files and other outputs remain outside the package license.

See the included LICENSE.txt for the complete terms.

Support

Support: haoyun18881@gmail.com

Product updates: https://x.com/BoltCat_AI

Use the support email for installation issues, reproducible bugs, documentation corrections, or purchase questions.