feat(storage): bind strict publication producers

[fishmingyu]

Summary

Bind strict publication producers to bounded metadata validation, callback-scoped directory readers, retained repository-source authority, and an exact trusted workspace-provider contract.

This is a stacked PR based on #590. It does not yet add the strict BM25 or context-artifact producer; those remain separate because they must reconcile the newer portable-inert native authorization contract.

Changes

• bound publishable JSON bytes, lexical tokens, nodes, depth, keys, strings, and atoms before DOM allocation; require UTF-8 without a BOM and validate case-insensitive JSON suffixes
• migrate static Wiki export to one OwnedDirectoryStage generation with authenticated frontend copying and exact staged/published reader validation
• retain one source-fingerprint-v2 binding across summaries, snippets, graph metadata, and source citations, then verify and close it before publication
• remove ambient Git-origin provenance from static export identities while preserving the interactive Wiki default
• bind strict workspace sessions to exact request, plan, destination expectation, PID, and one-shot provenance before any user callback or write
• serialize writes, validation, publication, revocation, and provider callback lifecycle through cancellation-safe gates
• preserve the first build or cancellation failure when stage/source cleanup also fails
• document the local compatibility boundary and the still-unwired strict provider milestone

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Refactoring
• Performance improvement
• Tests

Testing

• Tests pass locally
• Added new tests for the changes
• pytest -q test/test_atomic_directory.py test/test_captured_directory.py test/test_workspace_provider.py test/test_bounded_json.py test/artifacts/test_publish.py test/test_source_fingerprint.py test/test_repository_summary.py test/wiki/test_builder.py test/web/test_static_export.py test/web/test_repo_registry.py test/web/test_codemap.py test/graph/test_hierarchy.py --tb=short (569 passed, 15 skipped)
• independent frozen-diff adversarial review: 348 passed, 5 skipped; PID/concurrency/zero-publication matrix repeated 10 times (70 passed)
• unit tier excluding two known incomplete-litellm environment files: 4679 passed, 31 skipped; the 17 remaining failures all require attributes absent from this checkout's namespace-only litellm installation and are unrelated to this diff
• Black, isort with the Black profile, flake8, py_compile, and git diff --check on the changed surface

Checklist

• My code follows the project's style guidelines
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• My changes generate no new warnings
• Any dependent changes have been merged and published

Dependency: #590 is published, mergeable, and green but intentionally not merged yet. This PR remains draft until its updated remote checks complete.

[fishmingyu]

Temporarily returned to Draft after adversarial review found four release-blocking gaps: retained source authority for static export, pre-DOM JSON complexity bounds, workspace session revocation serialization, and destination expectation enforcement. Fixes and deterministic regressions are in progress. Dependent PR #592 is also Draft until this layer is revalidated.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2bc21fbf4e

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[fishmingyu]

The draft blockers from the previous review are now closed in f9f9fa8..b0b508e:

• JSON is complexity-bounded before DOM parsing and .JSON is validated.
• workspace callbacks are PID-bound, destination-expectation-bound, provenance-bound, and linearized with revocation.
• static source reads retain and reverify one source-v2 authority through generation; cleanup failures prevent publication.
• static repository identity no longer depends on ambient Git origin.

Independent final review of the frozen 20-file diff reported P0/P1/P2 = 0/0/0. The PR remains Draft while the new remote checks run and while #592 is restacked.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b0b508e836

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".