fix(deps): eliminate vulnerable zip 6 and 7 resolutions

[proerror77]

Change

Cargo Audit reports RUSTSEC-2026-0367, RUSTSEC-2026-0368, and RUSTSEC-2026-0369 for zip 6.0.0 and 7.2.0. All five affected lockfiles now resolve zip 9.0.5. The runtime lockfile contains no zip.

Upgrade direct archive dependencies and propagate zip 9 filename decoding failures. Keep Burn components at 0.20.1 and import the existing CPU components directly. This removes the unused Candle dependency from lock resolution. The burn-core alias retains the name required by Burn derive macros.

Prediction now uses the existing native DuckDB patch at 1.10504.0. This replaces upstream 1.10501.0, whose downloader required vulnerable zip. Arrow remains at its existing 58.3.0 resolution. The optional downloader also uses zip 9.0.5.

Issue relationship

None

Validation

• Actual six-workspace Cargo Audit: 0 vulnerabilities. No advisory ignores or audit bypasses added.
• Existing DuckDB modes contract: passed system/downloader compilation, missing-feature rejection, and bundled dependency isolation.
• Locked Cargo metadata for both changed workspace graphs: passed.
• Added archive rejection coverage for symlinks and malformed UTF-8 filenames.
• Archive component tests: 18 passed, including symlink and malformed UTF-8 rejection.
• Workspace ownership and dependency closure contracts: passed. CPU training requires the direct components and rejects the backend facade.
• Reviewed Linux default,db policy graph: all 174 packages and features unchanged; four input hashes and the graph identity refreshed.
• CPU research model tests: 20 passed. Prediction CPU model tests: 22 passed. Policy identity tests: 3 passed.
• Current-head Security CI and its actual six-workspace audit artifact: passed, 0 vulnerabilities.
• Current-head Monorepo CI, Prediction Markets CI, Security, and Docker Smoke: all passed. Claude draft review was skipped.
• git diff --check: passed.

Official advisories: 0367, 0368, 0369.

Runtime impact and rollback

Merged as 25f603cfb1e423567e3f2a8a80fc0491b4faf00d under explicit user authorization on 2026-10-19 at 14:06:52 UTC. All three automatic main checks passed. Deployment, cloud resource creation, and workflow re-enablement remain excluded. Reverting the fix restores the previous dependency graph, including vulnerable zip versions.

Post-merge verification

Merged main 25f603cfb1e423567e3f2a8a80fc0491b4faf00d passed Monorepo CI, Prediction Markets CI, and Security. The actual main audit artifact reports six workspaces and zero vulnerabilities.

The existing Release/GHCR automation succeeded using the verified Security CI image without rebuilding. Immutable digest readback passed. Both Alibaba-calling workflows remain disabled. No deployment, cloud resource creation, workflow re-enablement, or other PR merge was performed.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

• Configuration used: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: 6f3dea57-8343-462d-b639-3c03025ad65a

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-19T14:11:06.285810Z	862dfd5	Draft marked ready
🔒 Security Review	✅ Completed 2026-10-19T14:11:41.158234Z	862dfd5	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.