fix(errors): bound ValidationError rejected-value evidence

[seonghobae]

Closed unmerged — feasibility evidence for #132

Issue #132 identified a valid package-wide confidentiality footgun: protected-main ValidationError renders arbitrary rejected values with value!r and stores the original value in structured details by default. This Draft tested whether the generic constructor could be hardened immediately as a successor stacked on exact #105, which owns pg_llm_batch/exceptions.py.

Test-first experiment

• RED 5a49c3883e283860aa3329fa2a5cddbecf66d61e added focused confidentiality contracts requiring a non-disclosing default, hostile-__repr__ isolation, and explicit reviewed opt-in for non-sensitive value evidence.
• 7c910eeb2eb511f28753da5d6cc68aff7461b9f8 implemented that constructor policy on top of exact fix(errors): snapshot structured evidence inputs #105.
• 243ec4eeebb26209258c563e1fd276b91ea4b667 strengthened the disclosure-authority type boundary.
• f32f9217f5a2e71d938027ac2457926e28f7b2d8 exposed the integration problem: existing call-site contracts, such as provider endpoint validation, intentionally assert raw rejected-value evidence today.

Why this branch is not a feasible integration unit now

The generic default change is semantically package-wide. Making this successor green would require editing call-site tests and potentially production policy across surfaces currently owned by independent active PRs: #71 provider validation, #87 orchestrator/resource ownership, #104 request validation, and the #53 lifecycle stack. Continuing here would create a competing writer/integration branch rather than a narrow #105 successor. Those owners also need explicit classification of which rejected values are safe diagnostic evidence versus content-bearing/confidential values; blindly replacing every value with <redacted> would reduce useful diagnostics without a reviewed compatibility policy.

Therefore this experiment is closed unmerged. Its commits remain development evidence for #132, but no checks/reviews/approvals transfer and no behavior should be treated as shipped or as an accepted replacement for the active owners.

Next bounded path

Keep #132 open. After #105 and the relevant call-site owners reach protected main or are superseded, rebuild from the then-current protected result: establish the cross-package compatibility inventory first, classify each rejected-value surface, then apply the smallest safe constructor/call-site API with fresh RED→GREEN and exact-source evidence.

This closure is a failed-feasibility result, not a rejection of the privacy defect.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 4b6dc907-7da1-4950-bdd4-432d8d7aeed7

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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.}