astro-developer/examples/README.md

Independent Astro contributor experiment

This is a scoped synthetic experiment, not the Astro monorepo or an AI-client run. Tested Node 24.10.0, Astro 7.3.7 and Biome 2.5.3; the fixed skill snapshot is a separate repository identity. The Node type-erasure and Astro build APIs are experimental. Other tool versions have not been verified by this record.

From the extracted astro-developer directory, inspect sources and run:

npm ci --ignore-scripts --no-audit --no-fund
node examples/run_fixture.mjs

Installation contacts the free public npm registry; packages, binaries, node_modules and generated outputs are not distributed. The runner writes its own .generated directory and new examples/fixture-evidence.json. No preview server starts. Only synthetic local files are used. A preload denies JavaScript HTTP/socket connections during real standalone Astro builds and disables telemetry. No account, key, database, paid model or production data is used.

Forty-six observations include nine original blocks, actual Biome diagnostics, native original function/file effects and two Astro builds. The original Markdown unit assertion fails as expected due to a leading newline; the manifest unit assertion passes. A duplicate exported-function comparison is intentionally rejected. The overall BB harness treats these as explicit expected outcomes; it does not claim the original test pair is all green.

Other lint errors are retained separately from Node API rule results. Original generated bad code is inspected/linted, not executed. Output separation is not a reproduction of the monorepo loadFixture ESM-cache issue. No full TypeScript typecheck, browser/edge/adapter/pipeline matrix, authorization, speed, complete security or complete AI-client execution claim follows. Do not deploy the teaching fixtures. For real contributions obtain the complete matching Astro checkout. Original evaluation prompts are retained, not run.