Skip to content

About

Knowledge contribution prompts that willing teammates can run in their own approved AI tools.

Resources

Code of conduct

Contributing

Security policy

Stars

2 stars

Watchers

0 watching

Forks

Repository files navigation

Knowledge Contribution Kit

Large corpora provide breadth. The right teammates contribute judgment.

Why this exists

Large document collections can give an AI tool broad coverage, but they also contain substantial noise. The practical judgment that makes a skill or agent useful may sit with a small number of teammates who already do the work well, often in private day-to-day context that should not simply be handed to the builder.

Knowledge Contribution Kit gives builders one prompt shaped to the skill, agent, tool, checklist, or guide they are building. They can send it to willing teammates, who run it in their own approved AI environments and return one reviewed document of generalized guidance without handing over the underlying messages, email, or files.

Before inviting contributors, sending a contribution, or using workplace data, have a human lawyer at your organization review the process, contribution, intended data use, and sharing plan.

Use it

  1. Open the prompt builder.
  2. Describe what you are building and where the contributor's AI can read it.
  3. Generate and copy the contributor prompt.
  4. Send it to a willing teammate together with the project file or full text unless their AI can access the current version directly.
  5. The teammate runs it in their own approved AI environment, reviews the resulting document, obtains the required human legal review, and decides whether to send it back.

The page runs entirely in the browser. It does not upload or retain what the builder enters.

The latest release also works from downloaded local files without a server.

Why the prompt is shaped to what you are building

The generated prompt instructs the contributor's AI to:

  • read the current version of the skill, agent, tool, checklist, or guide before searching workplace context;
  • identify its existing domains, decision points, sources, boundaries, and output shape;
  • search only sources the contributor explicitly approves;
  • extract only the contributor's own knowledge;
  • keep only guidance that would materially improve the project;
  • compare every candidate with what the project already covers;
  • preserve exceptions, disagreement, uncertainty, and possible staleness;
  • remove identities, quotations, links, matter details, and confidential facts; and
  • produce one focused Word document with suggested improvements and insertion points.

The prompt has no arbitrary candidate cap. It consolidates overlap, ranks the five to ten strongest items in an executive summary, and includes every distinct candidate that would materially improve the project. If none qualify, it says so rather than padding the document.

After contributions come back

The page also generates a short synthesis prompt. The builder can use it to compare the returned contribution documents, identify overlap and disagreement, validate candidates against maintained sources, and prepare a change plan for the project being built.

Contributions are inputs for validation. Repetition does not turn a suggestion into policy or authority.

Limits

  • A prompt cannot guarantee that a contribution is de-identified or appropriate to share.
  • A hosted AI assistant processes information under the terms and settings that apply to that service. Confirm that the organization approves it for the proposed information.
  • Contributors should approve the source scope before any search begins.
  • Other people's statements may be used only for context and must not be extracted as the contributor's knowledge.
  • Nothing should be sent or changed automatically.

Development

Ordinary users do not need these commands.

npm install
npm test
npm run validate
npm run build

The project is distributed through GitHub and GitHub Releases, not npm.

License

MIT

About

Knowledge contribution prompts that willing teammates can run in their own approved AI tools.

Resources

Code of conduct

Contributing

Security policy

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages