Problem
The repo has no .github/ directory — zero CI, zero tests. Issues #13 and #14 (wrong throughput sort, wrong modality filter) have been shipping silently because nothing catches regressions before merge.
What CI should do
Frontmatter lint — verify every SKILL.md has required fields:
name: <non-empty string>
description: <non-empty string>
Script smoke test — for every script in */scripts/*.ts, verify:
npm install completes without error- Running the script with no args (or
--help) exits cleanly (not with a stack trace)
Optionally: a fixture test for list-models.ts and compare-models.ts that mocks the API response and asserts the sort output is correct — this would have caught #13 immediately.
Why now
The throughput sort bug (#13) is a correctness issue that produces silently wrong agent behavior. With no CI, similar bugs can land undetected. The fix is small — a single GitHub Actions workflow file.
Reviewed by Perry
Problem
The repo has no
.github/directory — zero CI, zero tests. Issues #13 and #14 (wrong throughput sort, wrong modality filter) have been shipping silently because nothing catches regressions before merge.What CI should do
Frontmatter lint — verify every
SKILL.mdhas required fields:Script smoke test — for every script in
*/scripts/*.ts, verify:npm installcompletes without error--help) exits cleanly (not with a stack trace)Optionally: a fixture test for
list-models.tsandcompare-models.tsthat mocks the API response and asserts the sort output is correct — this would have caught #13 immediately.Why now
The throughput sort bug (#13) is a correctness issue that produces silently wrong agent behavior. With no CI, similar bugs can land undetected. The fix is small — a single GitHub Actions workflow file.
Reviewed by Perry