Repository navigation
fix(security): clear SonarCloud new-code Security Rating - #30
Conversation
…rating Route remaining CLI-derived paths through validation and allowlist untrusted subprocess inputs so SonarCloud's new-code Security Rating returns to A: - mpfb_generate/run_catalog: resolve catalog + report paths via path_safety.resolve_under_roots (S8707) - check_feature_doc_matrix: confine changed-paths file under repo root and validate the git ref against a conservative pattern (S8707/S8705) - audit_dependencies: reject passthrough args with shell metacharacters (S8705) - check_doc_freshness: match link schemes without a clear-text URL literal (S5332) Dev-tooling hardening only; runtime behavior unchanged.
|
Warning Review limit reached
Next review available in: 15 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (9)
📝 WalkthroughWalkthroughNotebook embedded-visual assets now require explicit review fields. Developer tooling validates paths, refs, arguments, and URI schemes. CI uses locked ChangesNotebook review gate
Security and release hardening
Estimated code review effort: 3 (Moderate) | ~30 minutes Sequence Diagram(s)sequenceDiagram
participant ReleaseWorkflow
participant uv
participant PyPITrustedPublishing
ReleaseWorkflow->>uv: build source distribution and wheel
ReleaseWorkflow->>PyPITrustedPublishing: publish artifacts with GitHub OIDC
Possibly related PRs
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
There was a problem hiding this comment.
Actionable comments posted: 4
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@CHANGELOG.md`:
- Around line 15-22: Align the changelog entry with the checks actually enforced
by audit_dependencies: either implement validation for general shell
metacharacters on both the primary and pip-audit fallback argument paths, or
narrow the entry’s claim to the existing validation behavior. Update the wording
around “audit-arg allowlisting” and “rejects passthrough args with shell
metacharacters” without changing unrelated entries.
In `@scripts/audit_dependencies.py`:
- Around line 121-125: Centralize audit-argument validation in
scripts/audit_dependencies.py before selecting either uv or pip-audit, applying
an explicit policy for rejected inputs and emitting sanitized error messages;
update CHANGELOG.md lines 15-22 to document only the inputs actually rejected
unless broader hardening is added.
In `@scripts/check_feature_doc_matrix.py`:
- Around line 21-26: Sanitize validation exceptions in _resolve_within and the
additionally affected validation paths so they do not interpolate raw
CLI-supplied paths or Git refs. Replace values such as path and ref identifiers
in raised error messages with generic, constant descriptions while preserving
the existing validation and failure behavior.
In `@scripts/phantom_gen/mpfb_generate.py`:
- Around line 224-226: Validate args.catalog_id before constructing out_obj,
requiring it to be a single safe filename component via an explicit allowlist
that rejects path separators and traversal segments. Preserve the resolved
out_dir behavior, and fail clearly before any output file is created when the ID
is invalid.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: d5523e99-fe00-4d3a-994e-0875ca631faf
📒 Files selected for processing (6)
CHANGELOG.mdscripts/audit_dependencies.pyscripts/check_doc_freshness.pyscripts/check_feature_doc_matrix.pyscripts/phantom_gen/mpfb_generate.pyscripts/phantom_gen/run_catalog.py
…hing (OIDC) Replace the unpinned `pip install setuptools wheel twine build` + `python -m build` with `uv build` (pinned uv toolchain already installed in the job), clearing SonarCloud S8544/S8541 on the release path and dropping unused build tools. Switch the PyPI publish step to Trusted Publishing (OIDC): remove the stored PYPI_DEPLOY_API_KEY secret and add workflow `id-token: write`, so releases authenticate with a short-lived token instead of a long-lived key. The step stays inert until a GitHub Release is created and a PyPI trusted publisher is registered. Add PUBLISHING.md documenting the (optional) one-time PyPI setup; note the pipeline is inherited upstream cruft and unused by this fork.
… unpinned pip Convert the `build` matrix job and the main-only `cloud-scans-after-gates` job to `uv sync --extra dev --locked` + `uv run --no-sync` (cross-OS: no hardcoded .venv/bin vs .venv/Scripts), and replace `pip install codecov` with the pinned codecov/codecov-action. Clears SonarCloud S8544 (unpinned deps) on those jobs; safety/pytest/coverage now come from the locked env. The intentionally unpinned ci-latest.yml sweep is left as-is by design.
Automatic Analysis ignores sonar-project.properties, so add the file it does read (.sonarcloud.properties) with exclusions for data/asset dirs where private data is most likely to land (example_data, phantom_data, table_data, *.dcm, notebooks) plus build noise. Mirror the same exclusions into sonar-project.properties (local/CI sonar-scanner). Scan hygiene + defense-in-depth; not a privacy control.
There was a problem hiding this comment.
Actionable comments posted: 6
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.github/workflows/ci.yml:
- Around line 195-198: Update the “Install uv” setup-uv step to use the same
exact uv version and SHA256 checksum as the other setup-uv step, replacing the
range-based version input with matching version and checksum values while
preserving the existing action reference.
In @.github/workflows/release.yml:
- Around line 48-53: Extend the release workflow after the `uv build` and Grype
scan steps to generate and store an SBOM for `dist/`, attest build provenance,
sign every distribution artifact, and verify the signatures before the existing
publish step. Keep these controls as mandatory gates so publishing cannot
proceed when SBOM generation, attestation, signing, or verification fails.
- Around line 71-77: Correct the misleading pre-registration comment in the
“Publish distribution to PyPI (Trusted Publishing / OIDC)” step: either state
that tagged releases attempt publication and fail without a registered trusted
publisher, or add an explicit preflight condition that skips the step until
registration is complete. Keep the existing tag-based release behavior unless
implementing the gate.
- Line 12: Move id-token: write out of the workflow-wide permissions and into
the PyPI publish-only job. Split the production build and artifact-scanning
steps from publishing so the build job retains only the permissions it needs,
while the publish job explicitly grants id-token: write for Trusted Publishing.
In @.sonarcloud.properties:
- Around line 15-22: Remove the wildcard-based exclusion settings from
.sonarcloud.properties, including sonar.exclusions, sonar.test.exclusions, and
sonar.coverage.exclusions, and configure the intended patterns through SonarQube
Cloud Automatic Analysis or a CI/CD sonar-scanner/sonar-project.properties
configuration instead.
In `@PUBLISHING.md`:
- Around line 15-36: Update the Trusted Publishing setup guidance around the “no
PYPI_* secret” claim to include a one-time cleanup step: after verifying Trusted
Publishing works, delete the legacy PYPI_DEPLOY_API_KEY GitHub secret and revoke
its corresponding PyPI token. Make the documentation state that the no-secret
claim applies after this cleanup.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 374ee637-8c7d-4ad3-98cb-55d62fd57f53
📒 Files selected for processing (6)
.github/workflows/ci.yml.github/workflows/release.yml.sonarcloud.propertiesCHANGELOG.mdPUBLISHING.mdsonar-project.properties
| - name: Install uv | ||
| uses: astral-sh/setup-uv@11f9893b081a58869d3b5fccaea48c9e9e46f990 # v8.3.2 | ||
| with: | ||
| version: ">=0.11.19" |
There was a problem hiding this comment.
🔒 Security & Privacy | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "Workflow excerpt:"
if [ -f .github/workflows/ci.yml ]; then
nl -ba .github/workflows/ci.yml | sed -n '170,255p'
else
echo ".github/workflows/ci.yml not found"
fi
echo
echo "Relevant occurrences in workflows:"
rg -n "setup-uv|version:|pythonVersion|uv lock|uv lock --check|SBOM|provenance|cosign|sigstore|slsa|attest|dependency-audit|audit|SCA|SAST|zizmor|npm ci|npm install" .github/workflows || true
echo
echo "Lockfiles:"
git ls-files | rg '(^|/)(uv\.lock|uv.lock|requirements.*\.txt|poetry\.lock|pyproject\.toml|package-lock\.json|pnpm-lock\.yaml|yarn\.lock)$' || trueRepository: kgrizz-git/MyPySkinDose
Length of output: 219
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "GitHub Actions workflow files:"
find .github/workflows -maxdepth 1 -type f -print 2>/dev/null | sort || true
echo
echo "Workflow excerpt:"
if [ -f .github/workflows/ci.yml ]; then
awk '{printf "%4d\t%s\n", NR, $0}' .github/workflows/ci.yml | sed -n '170,255p'
else
echo ".github/workflows/ci.yml not found"
fi
echo
echo "Relevant occurrences in workflows:"
rg -n "setup-uv|version:|pythonVersion|uv lock|uv lock --check|SBOM|provenance|cosign|sigstore|slsa|attest|dependency-audit|audit|SCA|SAST|zizmor|npm ci|npm install" .github/workflows || true
echo
echo "Lockfiles and build config:"
git ls-files | rg '(^|/)(uv\.lock|uv.lock|requirements.*\.txt|poetry\.lock|pyproject\.toml|package-lock\.json|pnpm-lock\.yaml|yarn\.lock)$' || trueRepository: kgrizz-git/MyPySkinDose
Length of output: 7053
🌐 Web query:
astral-sh setup-uv version input pinned version checksum
💡 Result:
When using the astral-sh/setup-uv GitHub Action, you can specify both a pinned version of uv and a corresponding SHA256 checksum to ensure the security and determinism of your CI environment [1][2]. Action Configuration To pin a version and verify its integrity, use the version and checksum inputs in your workflow file [2][3]: - name: Install a specific version and validate the checksum uses: astral-sh/setup-uv@v9.0.0 with: version: "0.3.1" checksum: "e11b01402ab645392c7ad6044db63d37e4fd1e745e015306993b07695ea5f9f8" Version Management The action automatically verifies checksums for the default version it installs [2][4]. If you specify a version that is not yet known to the action, you must provide the checksum input to perform validation [2][4]. You can find the correct SHA256 hashes for any release on the official astral-sh/uv GitHub releases page [2]. Alternatively, you can rely on the action's built-in version resolution by pointing it to a configuration file, such as a uv.lock file or pyproject.toml, using the version-file input [1][5]. This is recommended for maintaining deterministic CI runs [1]. Security Best Practices For environments where you must avoid piping remote scripts directly into a shell (e.g., curl | sh), you can manually download a pinned release tarball from the astral-sh/uv releases page, verify its checksum using standard tools like sha256sum, and extract the binary to your desired path [6]. Official uv installer scripts now also support and embed checksum verification for increased security [7][8].
Citations:
- 1: https://github.com/astral-sh/setup-uv/blob/main/docs/advanced-version-configuration.md
- 2: https://github.com/astral-sh/setup-uv/blob/224c887d/docs/customization.md
- 3: https://deepwiki.com/astral-sh/setup-uv/4-configuration-reference
- 4: Supporting SHA in with.version field astral-sh/setup-uv#851
- 5: https://github.com/astral-sh/setup-uv/releases/tag/v8.3.0
- 6: Pin and verify the
uvinstall in CI instead ofcurl | shopenreview/openreview-expertise#371 - 7: Enable checksum verification in the generated installer script astral-sh/uv#18625
- 8: Verify checksum in installer script astral-sh/uv#13074
Pin the uv download to an exact release with checksum validation.
The CI job uses version: ">=0.11.19", so these workflow runs can resolve different uv binaries even though dependency installs are locked. Use the same exact uv version and SHA256 checksum in the other setup-uv step (lines 228-231); setup-uv supports version plus checksum inputs.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In @.github/workflows/ci.yml around lines 195 - 198, Update the “Install uv”
setup-uv step to use the same exact uv version and SHA256 checksum as the other
setup-uv step, replacing the range-based version input with matching version and
checksum values while preserving the existing action reference.
| permissions: | ||
| contents: read | ||
| actions: read | ||
| id-token: write # PyPI Trusted Publishing (OIDC): short-lived token, no stored secret |
There was a problem hiding this comment.
🔒 Security & Privacy | 🟠 Major | 🏗️ Heavy lift
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "Files:"
git ls-files .github/workflows 2>/dev/null || true
echo
echo "release.yml with line numbers:"
if [ -f .github/workflows/release.yml ]; then
nl -ba .github/workflows/release.yml | sed -n '1,140p'
fi
echo
echo "Security/distribution permissions and id-token references:"
rg -n "permissions:|id-token:|permissions:" .github/workflows/release.yml || trueRepository: kgrizz-git/MyPySkinDose
Length of output: 421
🏁 Script executed:
#!/bin/sh
set -euo pipefail
echo "release.yml with line numbers:"
awk '{printf "%6d\t%s\n", NR, $0}' .github/workflows/release.yml | sed -n '1,180p'
echo
echo "Relevant workflow sections:"
awk '{printf "%6d\t%s\n", NR, $0}' .github/workflows/release.yml | sed -n '1,35p;40,80p'Repository: kgrizz-git/MyPySkinDose
Length of output: 8063
🌐 Web query:
GitHub Actions workflow permissions jobs job-id override parent scope id-token write read
💡 Result:
In GitHub Actions, you can define permissions at either the workflow level (applying to all jobs) or the job level (applying only to that specific job) [1][2]. When you specify permissions within a specific job, it completely overrides any workflow-level permissions [2]. The job-level permissions do not merge with or inherit from the workflow-level settings; they replace them entirely [2]. Regarding the id-token permission: - The valid values for id-token are write and none [3]. - You cannot set id-token to read [3]. Attempting to do so is invalid as there are no id-token resources to read [3]. - Setting id-token: write is required to allow a job to request and generate an OpenID Connect (OIDC) JSON Web Token (JWT) for authentication with external services (such as cloud providers) [4][5]. This permission does not grant the ability to modify or write to repository resources [4]. Example of a job-level override: jobs: my-job: runs-on: ubuntu-latest permissions: id-token: write # Overrides any workflow-level permissions for this job steps: - name: Get OIDC Token run: echo "Requesting OIDC token"
Citations:
- 1: https://docs.github.com/en/actions/reference/workflows-and-actions/workflow-syntax
- 2: https://adaptive-enforcement-lab.com/secure/github-actions-security/token-permissions/job-scoping/
- 3: Wrong possible id-token permissions types listed github/docs#33483
- 4: https://docs.github.com/en/enterprise-cloud@latest/actions/how-tos/secure-your-work/security-harden-deployments/oidc-in-cloud-providers
- 5: https://github.blog/changelog/2023-06-15-github-actions-securing-openid-connect-oidc-token-permissions-in-reusable-workflows/
Move id-token: write to the publish-only job.
Currently the workflow-wide permission makes checkout, the release gate, uv build, and artifact-scanning steps eligible to mint the PyPI OIDC token. Split production build/scan from PyPI publishing and set permissions: ... id-token: write only on the publish job; keep the build job at least privilege.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In @.github/workflows/release.yml at line 12, Move id-token: write out of the
workflow-wide permissions and into the PyPI publish-only job. Split the
production build and artifact-scanning steps from publishing so the build job
retains only the permissions it needs, while the publish job explicitly grants
id-token: write for Trusted Publishing.
| # Build with the pinned uv toolchain already installed above (uv resolves | ||
| # build deps from the project) instead of an unpinned `pip install build`. | ||
| # Removes the SonarCloud S8544 (unpinned deps) finding on the release path and | ||
| # drops the unused setuptools/wheel/twine installs (gh-action-pypi-publish | ||
| # uploads on its own). | ||
| run: uv build |
There was a problem hiding this comment.
🔒 Security & Privacy | 🟠 Major | 🏗️ Heavy lift
Add release-artifact provenance controls before publishing.
The shown release path builds and Grype-scans dist/, but does not generate/store an SBOM, attest provenance, or sign and verify the distributions before upload. Add these gates before the publish job.
As per coding guidelines, CI/CD must generate SBOMs, attest provenance, and sign/verify artifacts before deployment.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In @.github/workflows/release.yml around lines 48 - 53, Extend the release
workflow after the `uv build` and Grype scan steps to generate and store an SBOM
for `dist/`, attest build provenance, sign every distribution artifact, and
verify the signatures before the existing publish step. Keep these controls as
mandatory gates so publishing cannot proceed when SBOM generation, attestation,
signing, or verification fails.
Source: Coding guidelines
Add dev-docs, *.log, *.txt, and image globs to the Sonar exclusion lists (both .sonarcloud.properties and sonar-project.properties). Align the stale local scanner file to projectKey/name kgrizz-git_MyPySkinDose / MyPySkinDose and add sonar.organization so it can actually run. Scan hygiene only; the commit/CI privacy gates remain the PHI/PII control.
…ok assets notebook_embedded_visual inventory entries now need a notebook_review block with embedded_images_reviewed + burned_in_text_reviewed both true; an approved status alone no longer clears a notebook with rendered image/PDF outputs (NOTEBOOK_REVIEW_FIELDS_INCOMPLETE). Mirrors the DICOM/container checklists so embedded outputs get an explicit human PII/PHI review. Updates the checker, inventory template, render checklist, the getting_started entry, docs, and tests.
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
tests/unittests/test_check_sensitive_content.py (1)
265-283: 🔒 Security & Privacy | 🔵 Trivial | ⚡ Quick winTest incomplete notebook checklist values.
These tests cover a missing
notebook_reviewobject and the all-true case, but not a present checklist with one flag set toFalseor a non-boolean value. Add negative cases to protect the strictis Trueadmission-gate contract.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/unittests/test_check_sensitive_content.py` around lines 265 - 283, Add negative tests alongside test_notebook_embedded_visual_requires_review_checklist and test_notebook_embedded_visual_with_completed_review_passes for a present notebook_review checklist containing a False flag and a non-boolean value. Assert both cases produce NOTEBOOK_REVIEW_FIELDS_INCOMPLETE, preserving the strict is True admission-gate behavior.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@dev-docs/approved_asset_inventory.md`:
- Line 24: Update the generated inventory instructions preceding the table to
reference notebook and container checklists alongside the DICOM checklist, then
regenerate dev-docs/approved_asset_inventory.md so its content reflects the
updated generator output.
---
Nitpick comments:
In `@tests/unittests/test_check_sensitive_content.py`:
- Around line 265-283: Add negative tests alongside
test_notebook_embedded_visual_requires_review_checklist and
test_notebook_embedded_visual_with_completed_review_passes for a present
notebook_review checklist containing a False flag and a non-boolean value.
Assert both cases produce NOTEBOOK_REVIEW_FIELDS_INCOMPLETE, preserving the
strict is True admission-gate behavior.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 9db085e5-8927-4518-b4d9-633e50751253
📒 Files selected for processing (7)
CHANGELOG.mddev-docs/PRIVACY_AND_SENSITIVE_ASSETS.mddev-docs/approved_asset_inventory.jsondev-docs/approved_asset_inventory.mdscripts/check_sensitive_content.pyscripts/render_asset_inventory.pytests/unittests/test_check_sensitive_content.py
🚧 Files skipped from review as they are similar to previous changes (1)
- CHANGELOG.md
- .sonarcloud.properties: wildcards are unsupported by Automatic Analysis, so drop the inert glob exclusions and point to the SonarCloud UI / sonar-project .properties (CLI) instead. - check_feature_doc_matrix: sanitize validation errors (no raw path/ref echoed). - mpfb_generate: validate catalog_id as a safe filename stem before use. - audit_dependencies + CHANGELOG: narrow wording to the control-char/whitespace check actually enforced on the shell-less uv audit call. - release.yml: scope id-token:write to the publish job; correct the publish-step comment (fails, not "inert", without a trusted publisher). - PUBLISHING.md: add one-time legacy-token delete/revoke step; afterward. - render_asset_inventory: intro now names DICOM/container/notebook checklists; regenerated the inventory Markdown.
…operties TEMPORARY experiment (to be reverted): add docs/sonar_probe/probe_control.py (not excluded) and probe_excluded.py (matched by `sonar.exclusions=**/probe_excluded.py`), each carrying a Sonar S1192 duplicated-literal signal. The PR's Automatic Analysis scans new files, so comparing whether the excluded probe is analyzed vs the control tells us empirically if wildcard exclusions in .sonarcloud.properties are honored.
…lusions The PR probe proved wildcard sonar.exclusions ARE honored by Automatic Analysis for this project (probe_excluded.py dropped, control analyzed), contradicting the docs. Remove the temporary probe files and restore the real glob exclusions in .sonarcloud.properties with a note on the doc-vs-observed-behavior gap.
|



Why
mainshows a red ❌ because SonarCloud's quality gate fails on Security Rating on New Code (required ≥ A). All GitHub Actions CI jobs are green; the red is purely SonarCloud. New-code analysis flagged MAJOR "vulnerabilities" from its new (2025) AI/taint rules.This PR fixes the Python/path/URL findings in code. The remaining ~21 findings are GitHub Actions
pip installrules (S8541unpinned/--only-binary,S8544) that are unavoidable for editable/dev installs — those are handled separately via SonarCloud triage (mark Accepted / disable rule).Changes (dev-tooling hardening only; runtime unchanged)
mpfb_generate/run_catalog— route catalog + JSON-report paths throughpath_safety.resolve_under_roots(S8707).check_feature_doc_matrix— confine the changed-paths file under the repo root and validate the--against-refgit ref against a conservative pattern (S8707 / S8705).audit_dependencies— reject passthrough audit args containing shell metacharacters before the (shell-less) subprocess call (S8705).check_doc_freshness— match external-link schemes without embedding a clear-texthttp://literal (S5332).Validation
ruff checkclean on all touched files.pytestgreen:test_phantom_gen_pipeline(18 passed), plustest_audit_dependencies/test_check_doc_freshness/test_check_feature_doc_matrix(31 passed).check_feature_doc_matrix.py --against-ref HEAD~1runs end-to-end.🤖 Generated with Claude Code
Summary by CodeRabbit
Bug Fixes / Security
Build & Release
uvinstalls and a pinned Codecov action; improved scan/tool execution.Documentation