What is this?
Entry
What does it cover that the current corpus does not?
xarray would add a substantial scientific-computing library with dense NumPy and pandas integration, public API re-exports, optional backends, dynamically registered accessors, decorators, overloads, and extensively typed code. Its cross-module imports and dependency-derived types make it a promising source of coverage for liveness and resolution behavior related to Skylos #735 and #737.
A repository-root scan would also exercise the optional ai-defects, danger, secrets, and quality analyses across a mature package, tests, documentation examples, benchmarks, and build configuration, providing several forms of rule coverage not represented by the smaller package-only entries in the current corpus.
Size and maintenance
GitHub reports a repository size of about 54 MB. The proposed pin was committed on 2026-08-18, and v2026.07.0 was released on 2026-07-09.
Two completed repository-root comparisons of Skylos 780432d231e3f92e7b143deff1d194df7da1b786 against 2ea6d056abaffaf640efc44acf3b3d5d297224a9, with every optional analysis enabled, produced 5,349 findings per side and no warnings. Measured cost was 127.47 to 128.35 seconds, with about 121 seconds end-to-end wall time on the benchmark host.
Anything else
Proposed Skylos target: repository root (.). Proposed corpus cost: 140.0 seconds, leaving modest headroom above the repeated measurements.
The root target is deliberate because the tests, documentation examples, benchmarks, and configuration add useful scanner coverage. If per-PR cost is too high, xarray/ is a plausible narrower target, but it should be benchmarked separately before assigning a lower cost.
What is this?
Entry
What does it cover that the current corpus does not?
xarray would add a substantial scientific-computing library with dense NumPy and pandas integration, public API re-exports, optional backends, dynamically registered accessors, decorators, overloads, and extensively typed code. Its cross-module imports and dependency-derived types make it a promising source of coverage for liveness and resolution behavior related to Skylos #735 and #737.
A repository-root scan would also exercise the optional ai-defects, danger, secrets, and quality analyses across a mature package, tests, documentation examples, benchmarks, and build configuration, providing several forms of rule coverage not represented by the smaller package-only entries in the current corpus.
Size and maintenance
GitHub reports a repository size of about 54 MB. The proposed pin was committed on 2026-08-18, and v2026.07.0 was released on 2026-07-09.
Two completed repository-root comparisons of Skylos 780432d231e3f92e7b143deff1d194df7da1b786 against 2ea6d056abaffaf640efc44acf3b3d5d297224a9, with every optional analysis enabled, produced 5,349 findings per side and no warnings. Measured cost was 127.47 to 128.35 seconds, with about 121 seconds end-to-end wall time on the benchmark host.
Anything else
Proposed Skylos target: repository root (.). Proposed corpus cost: 140.0 seconds, leaving modest headroom above the repeated measurements.
The root target is deliberate because the tests, documentation examples, benchmarks, and configuration add useful scanner coverage. If per-PR cost is too high, xarray/ is a plausible narrower target, but it should be benchmarked separately before assigning a lower cost.