Record multi-file external fixture evidence

[willamhou]

Summary

• Add fail-closed post-validation to dogfood external-fixture evidence and require post_validation_passed in external evidence verification.
• Teach explicit edit guardrails to process multiple replace ... with ... in ... requests in one task, and emit python3 in the multi-file fixture scaffold.
• Record verified online Python invoice multi-file external fixture evidence and update README/status/spec gap wording.

Validation

• cargo fmt --check
• git diff --check
• cargo test external_fixture --lib -- --test-threads=1
• cargo test multifile_edit --lib -- --test-threads=1
• cargo test derive_edit_requests_supports_multiple_file_replacements --lib -- --test-threads=1
• cargo test offline_planner_routes_git_blame_with_path_and_line --lib -- --test-threads=1
• cargo build
• scripts/create-multifile-external-fixture.sh /tmp/deepseek-external-fixtures/python-invoice-multifile-20260523-release
• target/debug/deepseek dogfood external-fixture --workdir /tmp/deepseek-external-fixtures/python-invoice-multifile-20260523-release --budget 12 --evidence-out .dscode/dogfood/external-fixture-python-invoice-multifile-evidence.json 'replace ... validate with python3 -m unittest discover -s tests'
• target/debug/deepseek dogfood external-evidence --file .dscode/dogfood/external-fixture-python-invoice-multifile-evidence.json --out .dscode/dogfood/external-fixture-python-invoice-multifile-verification.json --require-successful-external-fixtures 1 --json
• cargo test --lib -- --test-threads=1 (1635 passed)
• node scripts/check-secrets.js

[github-actions]

DeepSeekCode review of PR #16 (Record multi-file external fixture evidence)

Let me examine the PR diff by looking at the actual changes in the repository.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d3c17d8852

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".

[chatgpt-codex-connector]

Track pending explicit edits by replacement, not just file path

When multiple replace ... with ... in ... requests target the same file, the new pending-edit logic marks all of them complete after the first successful apply_patch, because edit_request_patch_succeeded only checks whether any successful patch summary contains the path. In that scenario the guardrail stops issuing subsequent required patches, so later replacements in the same file are silently skipped.

Useful? React with 👍 / 👎.