[corpus] Add fluids

[mcdigman]

What is this?

• Add a new project

Entry

• Project name: fluids
• Repository URL: https://github.com/CalebBell/fluids
• License (SPDX): MIT
• Proposed pin: 5fa218f9c15f62e2a7f864b97991f4837ddeefe7

What does it cover that the current corpus does not?

For Vulture, the fluids package provides a large natural example for jendrikseipp/vulture#421, which teaches Vulture about annotated and augmented __all__ declarations. The historical comparison drops 259 findings, from 659 to 400. This is substantially broader coverage than the small examples currently available.

For Skylos, scanning fluids with danger, secrets, quality, and ai-defects adds three rule IDs absent from the current all-modes corpus:

• SKY-D224 API-signature hallucination: 2 findings
• SKY-L001 mutable default argument: 11 findings
• SKY-P402 pandas read_csv without chunksize: 1 finding

The numerical/scientific code supplies different mutable-default shapes from the existing framework-heavy corpus. The pandas SKY-D224 findings also provide useful false-positive regression coverage.

Size and maintenance

GitHub reports about 8,489 KB. The project remains maintained; the proposed pin is from 2026-07-26.

Measured two-sided detector cost (--jobs 1, base plus head):

• Vulture fluids: about 0.95 seconds; proposed corpus cost: 2 seconds.
• Skylos fluids, dead code plus all four optional analyses: about 48.84 seconds and 1,985 findings per side; proposed corpus cost: 60 seconds.

Anything else

Suggested per-tool settings:

tools:
  vulture:
    targets:
      - fluids
    cost: 2.0
  skylos:
    analyses:
      - danger
      - secrets
      - quality
      - ai-defects
    targets:
      - fluids
    cost: 60.0