Skip to content

PRODUCT.md

Latest commit

 

History

History
59 lines (37 loc) · 3.53 KB

PRODUCT.md

File metadata and controls

59 lines (37 loc) · 3.53 KB
Raw

Product

Platform

web

Users

Developers and teams that use more than one AI coding CLI, or expect to change tools without rewriting project instructions and automation.

Product Purpose

agnostic-ai keeps project setup portable across AI coding tools. It also helps developers keep up with consequential vendor changes in an ecosystem that moves every week. Success means a team can maintain one setup, understand which upstream changes affect it, and adopt useful capabilities without binding the project to one tool.

Positioning

agnostic-ai combines native multi-target configuration emission with evidence-backed capability tracking. It connects upstream tool changes to the exact support, workaround, gap, or design decision in agnostic-ai.

Operating Context

Developers author canonical specs in a repository, run agnostic-ai sync, review generated native files, and use sync --check in CI. Maintainers audit vendor documentation and releases, file confirmed implementation work, and publish one developer briefing with each release on the repository's GitHub Pages site.

Capabilities and Constraints

  • The Go CLI supports 25 registered AI coding targets with different native formats and capability coverage.
  • GitHub Pages is a Zola static site built from docs/site/ and assembled with the browser playground.
  • User guides are canonical Markdown under docs/site/content/docs/. Zola publishes them at /docs/, and repository links point readers to the same public pages.
  • /agent-setup.txt gives coding agents one plain-text setup and migration contract generated from the canonical user guide. The public site and README expose the same ready-to-paste prompt.
  • CHANGELOG.md records released agnostic-ai behavior. Release briefings reproduce that record and separately cover verified upstream ecosystem changes, including changes agnostic-ai does not support yet.
  • Public claims need direct vendor evidence. Proposed shared capabilities are not presented as shipped support.
  • Updates are dated Markdown files with a generated archive and RSS feed. Readers can filter complete editions by target and search terms, with shareable URL state and a complete no-JavaScript archive. The site does not require a CMS or client-side framework.

Brand Commitments

Use the product name agnostic-ai. Keep the direct, technical, plain-language voice already used by the CLI and documentation. Preserve the existing site identity: warm neutral surfaces, rust accent, system sans typography, monospace labels, light and dark themes.

Evidence on Hand

  • Product and capability documentation under docs/site/content/docs/
  • Released changes in CHANGELOG.md
  • Existing GitHub Pages site in docs/site/
  • Evidence-backed target audit reports under the gitignored local/target-audit/ working directory
  • Confirmed findings and design decisions in labeled GitHub issues

Do not invent customer names, adoption figures, benchmarks, or commercial claims.

Product Principles

  • Keep project configuration portable without hiding native differences.
  • Explain the few upstream changes that affect developers now.
  • Separate observations, current support, proposed work, and released behavior.
  • Link every consequential claim to evidence and an actionable next step.
  • Prefer a small dependable release publishing flow over a broad content system.

Accessibility & Inclusion

The public site must remain usable with keyboard navigation, reduced motion, narrow screens, light and dark system themes, and semantic reading tools.