source-driven-development

Use source-driven development: detect exact versions from dependency files like package.json, fetch the relevant official docs page, implement the documented pattern, and cite the source URL. This prevents deprecated APIs from training data entering your codebase.

Identify the library version from your dependency file, then fetch the specific documentation page for that feature rather than the homepage. Extract API signatures, deprecation warnings, and migration notes, and implement exactly what the docs show.

Official documentation ranks first, followed by official blogs and changelogs, then web standards references like MDN, then compatibility data like caniuse.com. Stack Overflow answers, blog tutorials, and AI-generated summaries are never treated as primary sources.

The conflict is surfaced to the user with both options explained, such as using the modern documented pattern versus matching the existing codebase convention. The Skill never silently picks one approach without user input.

Skip it when correctness does not depend on a specific version, such as renaming variables, fixing typos, moving files, or writing pure logic like loops and data structures that behaves identically across versions.

Yes, fetched pages are treated as untrusted data. The Skill extracts only API definitions, examples, and deprecation notes while ignoring directives aimed at the model, and never hardcodes outbound endpoints from docs into generated code without surfacing them.