discussion: try to parse test output for each package (and mark individual test-cases to ignore)

I'm sure it's been brought up before, but I'm contemplating ways to parse the test output of each package and enable a finer grained ignore/flaky annotation, optimally on a per test case level, but on a suite level will be good as well.
My idea is to parse the package.json:scripts.test field, deduce test engine and args, and try to call engine directly with a TAP formatter. This could be a "best-effort" approach, with the current behaviour (npm test) as fall back.
AFAICT most packages use a well known test runner (mocha, tap, etc.) that know how to output TAP format.

Feedback is most appreciated...

Start by reviewing each package's package.json:scripts.test field and the existing npm test fallback. Investigate how common test engines can emit TAP, then define how per-test-case or suite-level ignore and flaky annotations would be represented. Done means best-effort parsing works where possible while npm test remains the fallback.