Description

Aligning outliers in os/* with namespace majority patterns (random namespace pick, seed 20260428).

Namespace summary

• Target namespace: @stdlib/os
• Members analyzed: 8 (arch, byte-order, configdir, float-word-order, homedir, num-cpus, platform, tmpdir) — all non-autogenerated
• Features analyzed: file tree, package.json shape, README headings, manifest.json shape, test/benchmark/examples filenames, public signature, return kind, validation prologue, error construction, JSDoc shape, @stdlib/* dependencies
• Features with clear majority (≥75%): universal file tree (13 files in 8/8), universal package.json top-level keys (19 keys in 8/8), ## Usage/## Examples/## CLI README headings (8/8), ## See Also (7/8), ## Notes (6/8), returnKind: value (8/8), hasExample: true (8/8)
• Features without clear majority: package.json browser key (3/8), manifest.json presence (2/8), benchmark/benchmark.js (3/8), errorConstruction (only configdir actually throws), publicSignature (7/8 export a constant; configdir exports a function with one optional parameter)

Per-package corrections

os/num-cpus

Populated the empty <section class="related"> block with a ## See Also listing for @stdlib/os/arch and @stdlib/os/platform, plus the corresponding <related-links> reference definitions. The section scaffold was already present but lacked all content. Brings num-cpus in line with 7 of 8 (87.5%) os/* siblings that carry a populated See Also list.

Validation

Checked:

• Structural feature extraction across all 8 members (file tree, package.json shape, README headings, manifest.json presence, test/benchmark/examples filenames).
• Semantic feature extraction via per-package agents (public signature, validation prologue, error construction, JSDoc shape, @stdlib/* dependency set).
• Three-agent drift validation (semantic-review, cross-reference, structural-review) on every flagged outlier.

Deliberately excluded:

• os/platform and os/tmpdir lacking ## Notes — both flagged by Agent 1 / Agent 3 as intentional deviations: platform is a thin pass-through over process.platform with nothing noteworthy beyond Node's own docs; tmpdir was split (Agent 1 leaned drift, Agent 3 leaned intentional, Agent 2 marked needs-human since meaningful Notes copy is editorial work). Per the routine, any needs-human or intentional-deviation verdict drops the outlier.
• Features without a clear ≥75% majority (browser key, manifest.json, benchmark/benchmark.js, error construction, signature shape).
• Outliers reflecting genuine semantic differences: C-binding extras in byte-order/float-word-order, lib/browser.js/lib/navigator.js in num-cpus/platform, configdir's optional path parameter and validation prologue.

The full per-feature majority/conformance/outlier table and the per-agent verdict matrix live in the run's local report.

Related Issues

None.

Questions

None.

Other

Generated by the cross-package API drift detection routine (majority-vote variant). Random namespace pick seed: 20260428. Eligibility filter: ≥8 non-autogenerated direct child packages. Conformance threshold: ≥75%. Validation pipeline: per-package semantic-feature extraction (sonnet) + three-agent drift validation (opus + opus + sonnet) with unanimous-non-rejection gating.

Checklist

• Read, understood, and followed the contributing guidelines.

AI Assistance

• Yes
• No

If you answered "yes" above, how did you use AI assistance?

• Code generation (e.g., when writing an implementation or fixing a bug)
• Test/benchmark generation
• Documentation (including examples)
• Research and understanding

Disclosure

This PR was authored by Claude (Opus 4.7) running the cross-package API drift detection routine. Namespace selection was deterministic from seed 20260428; structural feature extraction ran via shell tooling; per-package semantic feature extraction was delegated to parallel sonnet agents; the resulting outlier set was validated by three independent review agents (semantic-review, cross-reference, structural-review) before any change was applied. Only the single drift item that received a non-rejection verdict from all three reviewers advanced to a commit.

@stdlib-js/reviewers

Generated by Claude Code