Skip to content

PRD.md

Latest commit

 

History

History
110 lines (69 loc) · 5.7 KB

PRD.md

File metadata and controls

110 lines (69 loc) · 5.7 KB

Product Requirements Document: onData Skills

Overview

A curated collection of AI skills maintained by the onData association. The project aims to make powerful AI-assisted workflows accessible to everyone, with a focus on data work: reading, transforming, analyzing, and visualizing data. Skills are designed to be tool-agnostic: usable with any AI assistant or agent platform that supports the skill format.

Core Idea

A skill is, at its heart, a piece of structured knowledge: a description of how to approach a task, written so that an AI agent can follow it reliably. Writing a skill does not require programming ability. It requires knowing the task well — what it involves, what good output looks like, what can go wrong.

This makes skill creation a fundamentally inclusive activity. A data journalist who knows exactly how to audit an open dataset, a civic activist who has spent years navigating public records, a librarian who understands how metadata should be structured — any of them can write a useful skill, without writing a single line of code.

The technical layer (wrapping the skill in the right format, adding supporting scripts if needed) can come later, or can be handled by someone else. The knowledge layer — the hard part — comes from experience and context, not from developer skills.

This project is built on that premise: the people with the most useful knowledge are not always the people who write code, and the contribution model should reflect that.

Guiding Philosophy

The project is modeled on open source culture: a public repository, open contributions, transparent evolution, and community ownership. Skills are shared artifacts — anyone can use them, improve them, or propose new ones. The collection takes shape over time through collective effort, not top-down planning.

Problem Statement

The value is not in filling a gap — many skill collections already exist. The value is in doing it together: a group of people passionate about the same topics, pooling what they find useful in the hope that it will be useful to others too.

Goals

  • Provide a ready-to-use collection of skills that anyone can install and start using immediately.
  • Maintain a browsable catalog with clear titles and descriptions so people can find what they need.
  • Write everything in English to maximize reach and LLM comprehension.
  • Keep the barrier to entry as low as possible: simple installation, clear documentation, minimal prerequisites.

Scope

Skill Categories

There is no strict constraint on skill types. The collection will grow organically, but the core focus areas — reflecting the onData community's expertise — are:

  • Data reading — importing, parsing, and inspecting data from various sources and formats.
  • Data transformation — cleaning, reshaping, merging, and enriching datasets.
  • Data visualization — generating charts, maps, and visual summaries.
  • Data quality — validating, profiling, and auditing data.
  • Productivity & support — any skill that facilitates, supports, or improves workflows related to the above.

There are no thematic constraints. Any skill that has proven useful to someone in the community — regardless of topic — is welcome, on the assumption that it may be useful to others too.

What a Skill Contains

Each skill is a self-contained directory with:

  • A skill manifest (skill.md or equivalent) with metadata: name, description, category, usage instructions.
  • The skill prompt and any supporting files (templates, examples, reference docs).
  • A short entry in the project catalog for discovery.

User Experience

Discovery

A user should be able to:

  1. Open the catalog (initially a Markdown file in the repo).
  2. Browse skills by title and short description.
  3. Click through to a skill's directory for full documentation and usage examples.

Installation

A user should be able to:

  1. Clone or download the repository.
  2. Follow a simple README guide to register one or more skills in their AI tool of choice.
  3. Start using the skills immediately — no build steps, minimal external dependencies.

Phases

Phase 1 — Foundation

  • Define the project structure and skill format.
  • Write the README with installation instructions.
  • Create the catalog (Markdown-based).
  • Seed the collection with an initial set of skills.

Phase 2 — Web Catalog

  • Build a lightweight website that renders the catalog as a browsable, searchable interface.
  • Each skill gets its own page with full description, usage examples, and install instructions.
  • The site is generated from the same source data used by the Markdown catalog (single source of truth).

Phase 3 — Community Growth

  • Publish contribution guidelines.
  • Accept skill submissions via pull request: anyone can open a PR to add their own skill to the collection.
  • Introduce a review process to maintain quality.

Constraints

  • All content must be in English.
  • Skills must work without external dependencies unless strictly necessary.
  • The project structure must remain simple and navigable even without the web catalog.

Success Criteria

  • A non-technical user can install and use a skill within 5 minutes of reading the README.
  • The catalog provides enough information to choose a skill without reading its full source.
  • The collection covers at least the core data workflow (read → transform → visualize) at launch.

Open Questions

  • Should individual skills be versioned independently?
  • What is the curation process for community-submitted skills?
  • Should skills have maturity labels (e.g. experimental, stable)?
  • What technology should the Phase 2 website use (static site generator, GitHub Pages, etc.)?