Goal
Improve one published benchmark signal without widening the product scope. Choose exactly one:
- add one synthetic viewer-compatibility fixture and record the versioned result; or
- improve one deterministic rubric assertion with before/after raw evidence.
Allowed files
benchmarks/decision-v1/
storyboard_studio/benchmark.py
tests/test_benchmark.py
docs/BENCHMARK.md
Fixture and privacy boundary
Use synthetic or public data only, declare its license, and do not include private decks, customer content, API keys, or remote assets. Provider networking must remain opt-in.
Acceptance criteria
Maintainer availability
Ask contract questions here before implementing. A maintainer will confirm whether the proposed fixture stays inside the published rubric and privacy boundary.
Out of scope
Private data, new product templates, broad visual redesigns, mandatory cloud providers, and claims of factual or universal presentation quality.
Goal
Improve one published benchmark signal without widening the product scope. Choose exactly one:
Allowed files
benchmarks/decision-v1/storyboard_studio/benchmark.pytests/test_benchmark.pydocs/BENCHMARK.mdFixture and privacy boundary
Use synthetic or public data only, declare its license, and do not include private decks, customer content, API keys, or remote assets. Provider networking must remain opt-in.
Acceptance criteria
make benchmark-fixture-check,make lint,make format-check, andmake testpass.make benchmark-checkreports no unrelated score regression.score.jsonor a versioned viewer result is attached to the pull request.Maintainer availability
Ask contract questions here before implementing. A maintainer will confirm whether the proposed fixture stays inside the published rubric and privacy boundary.
Out of scope
Private data, new product templates, broad visual redesigns, mandatory cloud providers, and claims of factual or universal presentation quality.