SSA full drift checker (template + staging headers)

[mcxl]

Summary

Follow-up to #15. Promotes the minimal drift checker (1 check) to a full drift checker (3 checks) by first centralising the SSA report placeholder tokens.

• Refactor: build_ssa_report_docx's inline {{SITE_ADDRESS}} / {{NARRATIVE_SUMMARY}} / {{AUDIT_DATE}} / {{PREPARED_BY}} literals are now module-level PH_* constants plus an SSA_REPORT_PLACEHOLDERS tuple. Pure refactor — 114 docx-build tests unchanged.
• New check Claude/pims folder guide efy2 n #1 — template placeholders. Walks the SSA report template .docx's word/*.xml parts, concatenates <w:t> text content (so placeholders split across formatting boundaries are still caught), and asserts each PH_* appears.
• New check fix(pims): correct column name in observation fetch + drop debug wrapper #2 — staging xlsx widths. Every STAGING_HEADERS entry must have a corresponding _STAGING_COL_WIDTHS key. Subset semantics — _STAGING_COL_WIDTHS legitimately carries forward-defined gap-4/5/6 columns (phase, hrcw, swms_required, etc.) not yet in STAGING_HEADERS.

3 checks total, all green on current branch. 5 drift-checker tests cover the clean path and per-check failure paths.

Test plan

• python tools/check_ssa_drift.py — exits 0 with 3 [X] marks
• python -m pytest tests/test_ssa_drift_checker.py — 5 tests pass
• python -m pytest tests/test_ssa_pipeline.py tests/test_ssa_determinism.py tests/test_ssa_oxml_validator.py — docx-build behaviour unchanged after the placeholder refactor (114 passing)

Notes for reviewer

• Stacked on top of SSA pipeline quality hardening (Phases 1-6) #15. GitHub will auto-rebase the base when SSA pipeline quality hardening (Phases 1-6) #15 merges.
• The placeholder check uses a <w:t>...</w:t> regex to extract Word's visible text content rather than substring-searching the raw XML, because Word may split a placeholder across multiple text runs when formatting boundaries land mid-token.
• The staging-width check is deliberately a subset (not equality) — _STAGING_COL_WIDTHS has 6+ forward-defined gap-4/5/6 entries not yet in STAGING_HEADERS. Asymmetric semantics catch the case that matters (missing widths) without requiring those forward-defined entries to be removed.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d04355855b

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]

Reject split placeholders that the builder cannot replace

When Word splits a placeholder across formatting runs, this join makes the drift check pass because {{SITE_ + ADDRESS}} is found in the concatenated haystack, but build_ssa_report_docx still replaces only tokens that appear wholly inside one <w:t> (_replace_tokens_in_part iterates individual text nodes). In that edited-template scenario the checker reports green while the generated report keeps unreplaced placeholder pieces, so the check should either require each placeholder within a single text node or the builder needs matching split-run replacement support.

Useful? React with 👍 / 👎.