[justaugustus]

What kind of change does this PR introduce?

Refactor with tests (infra/tests/other). No change to scores, findings, or output.

• PR title follows the guidelines defined in our pull request documentation

What is the current behavior?

Every check repeats the same body: collect raw data, store it in RawResults, run the probes, evaluate, and wrap
errors at two points. Probe mode (--probes) wires the same checks to the same collectors a second time, in the
138-line assignRawData switch in pkg/scorecard. That switch carries a nolint:gocognit,gocyclo and had almost no
unit coverage. Adding a check means wiring it in both places.

What is the new behavior (if this is a feature change)?

Behavior is unchanged. Structurally:

• An unexported definition in checks pairs a check's collector with the RawResults field it fills, its probes,
  and its evaluation. define[T] types the collector and the field together, so pairing a collector with the wrong
  field fails to compile.
• runDefinition is the one copy of the collect, probe, and evaluate body.
• Maintained (a CheckRequest collector) and CI-Tests (a RepoClient collector, through an adapter closure) run
  from definitions. The exported checks.Maintained and checks.CITests keep their signatures and become one-line
  wrappers.

This is a pilot for two checks, sized so the shape can be judged before anything else moves. It is the first step
toward one check table that both check execution and probe mode read, which would let assignRawData go.

• Tests for the changes have been added (for bug fixes/features)

Two commits, each reviewable alone:

1. Characterization tests for probe mode (populateRawResults and assignRawData, for both checks). They pass on
   the current code and pin which collectors run, RepoClient call counts, stored raw data, per-check
   deduplication, stopping at the first error, and exact error text.
2. The migration, plus differential tests. The previous check bodies are kept verbatim in the test file. Each
   scenario runs the old body and the definition on identical inputs and requires identical results, details, raw
   data, and RepoClient call sequences, with and without a shared RawResults.

Which issue(s) this PR fixes

NONE

Special notes for your reviewer

Main question: does the definition read better than the body it replaces? Compare checks/ci_tests.go before and
after. If it doesn't, this should stop here, and I'd rather hear that now than after more checks move. The two
checks add net lines, because checks/definition.go is a one-time cost. Code only shrinks once most checks migrate
and probe mode reads the definitions.

Out of scope, deliberately:

• Registration stays in each check's init(). Probe mode still uses assignRawData for these two checks.
• Making probe mode read the definitions would need an exported accessor in checks and would change how
  check-registry is implemented, so it would come with an OpenSpec change first.

Two existing behaviors the tests surfaced, left unchanged because changing them would alter output:

• raw.CITests collates through a map, so CIInfo, findings, and details are in random order when a repo has more
  than one merged PR. The tests sort before comparing.
• evaluation.CITests ignores its name argument on success paths and uses its own CheckCITests constant. The
  check name is defined in checks, checks/evaluation, and internal/checknames.

Verification beyond CI:

• SKIP_GINKGO=1 go test -race -count=1 ./... passed locally.
• Mutations were applied one at a time and reverted. Each one fails the new tests: dropping the probe-mode seen
  update, dropping the CITestResults assignment, storing raw data when the collector fails, and giving CI-Tests the
  wrong check name.

Rollback: revert the second commit. The first commit's tests pass on the current code either way.

Does this PR introduce a user-facing change?

NONE

[codecov]

Codecov Report

❌ Patch coverage is 92.30769% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.68%. Comparing base (353ed60) to head (a121e1e).
⚠️ Report is 466 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #5425       +/-   ##
===========================================
+ Coverage   66.80%   79.68%   +12.87%     
===========================================
  Files         230      263       +33     
  Lines       16602    14897     -1705     
===========================================
+ Hits        11091    11870      +779     
+ Misses       4808     3026     -1782     
+ Partials      703        1      -702     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.