project-delivery-packager

Experimental skill for preparing, organizing, or reviewing a software, data, AI, RAG, knowledge-base, API, deployment, client handoff, acceptance, or internal project delivery package. Use for produci

SKILL.md

GitHub

Project Delivery Packager

Use this skill to turn a working project into a handoff-ready delivery package. The package should help another person deploy, test, understand, accept, and maintain the system without relying on private chat history.

Status: experimental. For the first real use on a new project type, produce a delivery plan or audit first and ask for confirmation before restructuring files, creating many docs, or packaging for external sharing.

Core Workflow

  1. Determine delivery type:

    • code package, deployment package, API service, data pipeline, RAG/knowledge-base system, model demo, acceptance package, or project archive.
    • identify recipient: client, teammate, teacher, ops engineer, reviewer, or future self.
  2. Inspect existing artifacts:

    • README, scripts, configs, API routes, database/schema files, test files, sample data, generated outputs, docs, and deployment notes.
    • do not expose secrets, private endpoints, tokens, client-sensitive data, or raw private datasets.
  3. Create the delivery map:

    • read references/delivery-structure.md before adding files.
    • separate source code, docs, examples, tests, deployment files, and generated outputs.
  4. Write handoff docs:

    • README, quick start, environment variables, install steps, run commands, API reference, data contract, database notes, and troubleshooting.
    • read references/doc-templates.md when drafting docs.
  5. Add acceptance materials:

    • test command list, smoke-test plan, sample input/output, known limitations, delivery inventory, and change log.
    • read references/handoff-qa.md before final delivery.
  6. Verify:

    • run lightweight commands when feasible.
    • if tests cannot run, explain why and provide manual verification steps.

Output Modes

  • delivery-plan: outline the delivery folder and required docs without editing files.
  • build-delivery-pack: create or update docs and package structure.
  • review-delivery-pack: audit an existing delivery package for missing handoff material.
  • acceptance-pack: produce acceptance checklist, test commands, sample cases, and delivery inventory.
  • sanitize-for-sharing: identify private details that must be removed before public or client sharing.

Default Delivery Folder

delivery_package/
  README.md
  docs/
    deployment.md
    api.md
    data_contract.md
    database.md
    test_commands.md
    troubleshooting.md
  examples/
    sample_input/
    sample_output/
  scripts/
  tests/
  delivery_inventory.md
  acceptance_checklist.md

Use this shape when creating a standalone delivery package. For an existing repo, add only missing docs and avoid disruptive restructuring.

Privacy And Client-Safety Rules

  • Never include real API keys, cookies, passwords, internal tokens, private hostnames, or personal data.
  • Replace sensitive values with placeholders such as YOUR_API_KEY, HOSTNAME, or CLIENT_SAMPLE_ID.
  • Do not publish raw client data. Use sanitized samples or synthetic examples.
  • When the user asks to upload, publish, or share a delivery package, stop and confirm the exact target and privacy boundary first.

Final Response Contract

When delivering, report:

  • package path or changed files;
  • what a recipient should run first;
  • what tests or smoke checks were performed;
  • what remains manual or environment-dependent;
  • any red flags about secrets, missing data, or unclear acceptance criteria.