Repo improvement sweep, 2026-08-17. Findings below are filed as separate issues and linked back here.
Profile
mature — 695 commits, created 2024-04-14, 45 contributors, 30 published releases, 91 closed bug issues, 51.8k lines of Python across 283 files with 151 test files. This repo got the full mature lens set, unlike the five greenfield repos swept alongside it.
Findings
Lenses run
repo-audit, repo-hygiene-audit, audit-code, improve-codebase-architecture
Lenses skipped
| Lens |
Reason |
fallow-risk-cleanup |
TypeScript/JS only — this is a Python repo |
closed-bug-regression-audit |
Deferred, not skipped for lack of signal: 91 closed bugs is the largest regression-audit surface of the six repos and deserves its own pass rather than a corner of a config sweep |
repo-doc-audit |
It rewrites docs — belongs in its own PR, not a filing sweep |
Recorded here rather than only in chat: six months from now this is the only record of whether a gap was checked-and-clear or never looked at.
Deliberately not filed
yaml.load at command/generate/spec.py:288,520 is a false positive. Ruff's S506 flags _YamlLoader as unsafe, but it subclasses yaml.SafeLoader and only strips YAML 1.1's on/off/yes/no bool aliases. Not a deserialization risk. Deliberately not filed.
S105 hardcoded-password-string (7 hits) are variable-name matches, not credentials. No committed .env or key material found.
- CI coverage is good — pytest, ruff, build-and-test, mac/windows/GPU matrices, CLA, AGENTS.md integrity. The gap is that none of it is required, which is filed separately.
- Issue templates, Dependabot config, pre-commit hooks,
AGENTS.md, CLAUDE.md, LICENSE and repo topics are all already present — not filed.
Filed by the repo-improvement-sweep skill.
Repo improvement sweep, 2026-08-17. Findings below are filed as separate issues and linked back here.
Profile
mature — 695 commits, created 2024-04-14, 45 contributors, 30 published releases, 91 closed bug issues, 51.8k lines of Python across 283 files with 151 test files. This repo got the full mature lens set, unlike the five greenfield repos swept alongside it.
Findings
criticaltrivialtar.extractall()runs with the extraction filter disabled on both paths #725 —tar.extractall()runs with the extraction filter disabled on both pathsrecommendedmediumexecute(complexity 48) andexecute_cloud(44) incommand/run/__init__.py#726 —execute(complexity 48) andexecute_cloud(44) incommand/run/__init__.pyrecommendedlargeBandSare off and surface 350+ real hits #727 — Widen the ruff ruleset —BandSare off and surface 350+ real hitsrecommendedmediumrecommendedsmallrecommendedsmallnice-to-havesmallLenses run
repo-audit,repo-hygiene-audit,audit-code,improve-codebase-architectureLenses skipped
fallow-risk-cleanupclosed-bug-regression-auditrepo-doc-auditRecorded here rather than only in chat: six months from now this is the only record of whether a gap was checked-and-clear or never looked at.
Deliberately not filed
yaml.loadatcommand/generate/spec.py:288,520is a false positive. Ruff'sS506flags_YamlLoaderas unsafe, but it subclassesyaml.SafeLoaderand only strips YAML 1.1'son/off/yes/nobool aliases. Not a deserialization risk. Deliberately not filed.S105 hardcoded-password-string(7 hits) are variable-name matches, not credentials. No committed.envor key material found.AGENTS.md,CLAUDE.md, LICENSE and repo topics are all already present — not filed.Filed by the
repo-improvement-sweepskill.