Restore Report Voice While Keeping Source-Correct Corrections

Summary

Rework the corrected reports so they read like technical comparison reports again, not audit memos. The implementation should preserve the current source-backed factual corrections, but remove rubric-like phrasing such as "this correction pass," "the reviewed source set does not support," and similar evidentiary narration unless a narrow qualification is genuinely necessary for accuracy.

Target the following report set as one coordinated style-restoration pass:

• docs/reports/1_Regulatory_Audit_Report.md
• docs/reports/2_Planning_Content_Report.md
• docs/reports/3_Engineering_Content_Report.md
• docs/reports/4_Maintenance_Protocol_Report.md
• docs/reports/8_Master2026_Content_Report.md

Use the corresponding backup files in .agents/backup/Reports-root-legacy-2026-04-09/ as the tone/template reference, especially for report 8. Keep the corrected facts, narrower claims, and valid citations from the current versions; restore the narrative report voice from the backup versions.

Key Changes

Report-writing posture

• Rewrite each target markdown report into backup-style narrative prose.
• Keep the existing section structures already associated with each report; do not turn them into grading rubrics or correction logs.
• Integrate factual corrections silently into the body text wherever possible.
• Use citations/source references as support, not as meta-commentary about the correction process.
• Remove recurring audit phrases such as:
• "the source set does not support"
• "this correction pass"
• "supported elsewhere in the corpus" when it can be replaced by direct attribution in normal prose
• "the defensible takeaway is"
• Keep a claim-qualification sentence only where needed to avoid repeating a previously unsupported statement.

Source and citation handling

• Re-check the current corrected claims against Sources before rewriting them into narrative form.
• Preserve narrowed or removed claims where the correction was substantively necessary.
• If a point depends on broader-corpus support, cite or attribute it in normal report prose instead of presenting it like a grading note.
• Prefer direct source-backed statements over meta-language about what was removed.
• For report 8, use the legacy backup as the style baseline and the current corrected version as the factual baseline.

Per-report implementation standard

• Report 1: restore an executive-summary and rule-comparison voice rather than a source-audit voice.
• Report 2: restore planning-report tone, with stable narrative treatment of continuity vs. 2026 changes.
• Report 3: restore engineering comparison tone while keeping the corrected rule/manual distinctions.
• Report 4: restore maintenance-report tone and keep fallback practice-specific details framed as examples or cited supporting material, not universal Chapter 7/8 holdings.
• Report 8: explicitly shift back toward the backup synthesis/report style while retaining the corrected Chapter 12/13/14 and NJAC substance.

HTML and publication parity

• Update the standalone HTML companions for the same five reports after markdown is locked:
• 1_Regulatory_Audit_Render.html
• 2_Planning_Content_Render.html
• 3_Engineering_Content_Render.html
• 4_Maintenance_Protocol_Render.html
• 8_Master2026_Content_Render.html
• Preserve current layout/presentation.
• Bring visible prose into parity with the rewritten markdown.
• Clean touched mojibake while editing.
• Regenerate the matching site/reports markdown, standalone HTML, and built page outputs with python scripts/build_green_guides.py.

Public Interfaces And File Contracts

• Canonical authored sources remain the docs/reports/*.md and docs/reports/*_Render.html files.
• Generated outputs remain under site/reports/.
• No section re-scoping or report-merging: each report keeps its existing purpose and section layout.
• The backup files are style references only; they do not override corrected source-backed substance.

Test Plan

• Confirm each rewritten markdown report reads as a narrative technical report, not a correction memo.
• Confirm each report still preserves the current corrected factual posture against Sources.
• Spot-check at least several previously corrected claims in each report to ensure the style restoration did not reintroduce unsupported statements.
• Verify that broad-corpus support, where still needed, appears as normal attribution rather than rubric-style fallback language.
• Compare report 8 against its backup to confirm the tone is materially restored while the corrected substance remains intact.
• Verify each standalone HTML file matches the rewritten markdown on all touched passages.
• Verify touched mojibake is removed.
• Run the normal site build and confirm the regenerated site/reports outputs reflect the rewritten docs-side sources.

Assumptions And Defaults

• This is a style-restoration correction pass, not a substantive rollback.
• Current source-backed factual corrections remain in force unless a fresh source review shows one should be adjusted again.
• The default rewrite posture is backup-style narrative, not hybrid audit prose.
• HTML parity and regenerated site/reports outputs are part of the same implementation pass even where the user only named the markdown files.

Part 2

Correct Phase 2B Small-Scale GI Report While Preserving Narrative Voice

Summary

Perform a source-first correction pass on the Phase 2B small-scale GI report, using the original OPAL prompts as the governing structure and the backup report as the tone reference. The implementation should keep the current five-section report structure, preserve the report-style narrative voice, correct unsupported or overstated claims against Sources, update the standalone HTML to factual parity, clean touched mojibake, and regenerate the corresponding site/reports outputs.

Canonical authored targets: - docs/reports/5_SmallScaleGI_Content_Report.md - docs/reports/5_SmallScaleGI_Content_Render.html

Generated outputs to refresh after docs-side corrections: - site/reports/5_SmallScaleGI_Content_Report.md - site/reports/5_SmallScaleGI_Content_Render.html - site/reports/5_SmallScaleGI_Content_Report/index.html

Key Changes

Source validation and writing posture

• Use docs/tools/workflows/Opal/Phase_2/2B_GeneratePhase2BSmallScaleGIContent.md as the structure contract and subject-matter scope.
• Treat the primary authorities as the 2023 and 2026 Chapter 9 small-scale GI chapters plus the 2023/2026 GI requirement fact sheets.
• Use broader Sources only when needed to support or narrow a claim, but keep the prose report-like rather than audit-like.
• Use the legacy backup report as a voice reference only; retain current source-correct substance, not legacy unsupported statements.
• Remove correction-memo phrasing and rewrite into narrative comparison prose similar to reports 1-4 and 8.

Section-level corrections

• Section 1 should frame Chapter 9 as the small-scale GI chapter in both editions and describe the BMP families actually present in the source set, without claiming that 2026 broadly moved Chapter 9 content out of scope.
• Section 2 should correct BMP-design statements to match chapter-specific source text, especially:
• keep small-scale bioretention as a broader family that includes rain gardens, stormwater planters, islands, street trenches, and enhanced/continuous tree pits where the chapter says so
• verify and correct drainage-area limits BMP by BMP rather than using blanket thresholds
• distinguish infiltrating and underdrained configurations where the chapters do
• remove or narrow unsupported universal design thresholds
• Section 3 should present treatment and performance in source-backed terms rather than invented universal metrics or generalized VRC-style rules.
• Section 4 should recast 2023 vs 2026 changes around actual reorganization and clarification in the sources:
• remove the unsupported storyline that 2026 Chapter 9.5 and 9.9 were moved out of Chapter 9
• instead state the supported distinction that GI-qualified MTD content remains in Chapter 9.5 while non-GI MTDs are addressed elsewhere, and that Chapter 9.9 addresses the GI-qualifying small-scale sand filter treatment while non-GI underdrained variants are excluded from that chapter treatment
• verify any claim about small-scale versus large-scale bioretention thresholds before retaining it; if the exact split is not clearly supported, replace it with narrower source-backed language
• Section 5 should restore practical report voice while keeping only source-backed implications for designers and reviewers, such as BMP-specific sizing limits, pretreatment expectations, underdrain implications, reuse considerations, and GI-distribution concepts from the fact sheets.

HTML parity and presentation

• Update docs/reports/5_SmallScaleGI_Content_Render.html only after the markdown is locked.
• Preserve the current standalone HTML layout, section structure, navigation, and visual presentation model.
• Bring all visible report prose into factual parity with the corrected markdown.
• Clean visible mojibake in touched content, including title/nav/article text and any damaged symbols that appear in human-facing copy.
• Keep the standalone HTML as a maintained artifact rather than replacing it with a redirect.

Public Interfaces And Contracts

• Canonical authored report remains docs/reports/5_SmallScaleGI_Content_Report.md.
• Canonical standalone HTML remains docs/reports/5_SmallScaleGI_Content_Render.html.
• Generated site/reports/... files remain output-only artifacts.
• The report must keep the existing five-section Phase 2B structure established by the OPAL generation prompt.
• Source references should support the narrative naturally; the report should not read like a grading rubric or correction log.

Test Plan

• Confirm the corrected markdown still follows the five intended sections from the Phase 2B prompt.
• Confirm the rewritten markdown reads like a technical comparison report, not an audit memo.
• Spot-check corrected claims against the chapter sources and fact sheets, including:
• Chapter 9.5 and 9.9 remaining present in the 2026 Chapter 9 source set
• small-scale bioretention variants such as rain gardens and related forms being treated as part of the family where the chapter says so
• BMP-specific drainage-area limits from the chapter tables and GI fact sheet
• underdrained versus infiltrating distinctions for small-scale GI BMPs such as pervious paving and bioretention
• any retained comparison claim about 2023 to 2026 chapter reorganization
• Verify that unsupported universal thresholds, performance percentages, or documentation mandates are removed or materially narrowed.
• Verify the standalone HTML matches the corrected markdown on all touched claims and that touched mojibake is removed.
• Run python scripts/build_green_guides.py and confirm the regenerated site/reports outputs reflect the corrected docs-side files.

Assumptions And Defaults

• This is a correction-and-voice-restoration pass, not a rewrite of the report’s scope or structure.
• Primary authority remains the 2023/2026 Chapter 9 family and GI fact sheets; broader Sources may be used only to support or narrow claims where needed.
• The backup report is a style reference, not a factual authority.
• If a current claim cannot be supported clearly from the source corpus, it should be deleted or restated more narrowly rather than preserved.

Part 3

Add Standardized Source Documents Blocks to Green Guides Reports

Summary

Add a consistent **Source Documents:** block to every canonical report in docs/reports, covering all *_Report.md files and keeping the standalone *_Render.html files in visible parity where they exist. Standardize existing **Source Materials:** labels to **Source Documents:**, exclude non-report support docs like index.md and 20260408_voice-correction.md, and use report-specific hyperlinks into the existing docs/Sources corpus.

Key Changes

• Target every report markdown in docs/reports matching *_Report.md, including 9_Validation_Report.md.
• Target every standalone report HTML in docs/reports matching *_Render.html so the rendered report pages visibly show the same source-document information.
• Standardize the label to exactly **Source Documents:** in markdown; remove **Source Materials:** wording where it exists.
• Place the new block in the same metadata zone near the top of each report, adjacent to the existing date / prepared-for / phase header area, so all reports share one consistent header structure.
• Use report-specific links into docs/Sources, preferring the markdown twins under docs/Sources/*/Markdown/*.md when available; link to the paired PDF only if a markdown twin is missing.
• For reports with a small, bounded source set, list the exact source documents directly.
• For reports with broader source coverage, use grouped, human-readable source summaries with direct links to the exact key files that define the report scope, rather than a generic link to sources-readme.md.
• Preserve report voice and body content; this pass is metadata/header augmentation, not content rewriting.
• Keep HTML layout intact; add a compact visible source-documents block near the top of each standalone render rather than changing page structure.
• Regenerate the corresponding site/reports outputs after docs-side updates so the published report pages reflect the new source-document blocks.

Public Interfaces And Content Contracts

• Canonical authored files remain the docs-side report sources under docs/reports.
• docs/Sources becomes the hyperlink target for report source provenance.
• No changes to report section structure or substantive narrative content.
• No changes to support docs in docs/reports that are not canonical reports.

Test Plan

• Verify every docs/reports/*_Report.md file contains a **Source Documents:** block.
• Verify reports that previously used **Source Materials:** now use **Source Documents:**.
• Verify every linked source path resolves within docs/Sources and uses the intended markdown twin when present.
• Verify every docs/reports/*_Render.html file displays matching source-document text and links.
• Spot-check representative reports across the set:
• report 1 and 2 preserve existing source-doc style while gaining links where needed
• reports 5-8 convert from Source Materials to linked Source Documents
• report 3 and 4 gain a new standardized block without disturbing their restored report voice
• report 9 gains a source-documents block even though it has no standalone render
• Run the normal site build and confirm the regenerated site/reports pages include the new source-document blocks and that no links are broken by the added docs/Sources references.

Assumptions And Defaults

• “Each report in docs/reports” means canonical report markdown files plus matching standalone HTML renders, not support pages like index.md or 20260408_voice-correction.md.
• The link target should be the existing docs-side source mirror under docs/Sources, not the repo-root Sources directory.
• Report-specific source links should point to exact relevant source documents, not a single shared source index page.
• This is a standardization and provenance pass only; report body substance and tone should remain otherwise unchanged.

Report 7

Correct Phase 2D Non-GI / MTD Report Against docs/Sources While Preserving Narrative Voice

Summary

Perform a source-first correction pass on the Phase 2D report using the original OPAL prompts as the structure contract and the restored-report style from the prior report passes as the voice target. The implementation should keep the current report cadence, preserve narrative comparison voice, correct unsupported or overstated claims against docs/Sources, update the standalone HTML to factual parity, clean touched mojibake, and regenerate the corresponding site/reports outputs.

Canonical authored targets: - docs/reports/7_NonGI_MTD_Content_Report.md - docs/reports/7_NonGI_MTD_Content_Render.html

Generated outputs to refresh after docs-side corrections: - site/reports/7_NonGI_MTD_Content_Report.md - site/reports/7_NonGI_MTD_Content_Render.html - site/reports/7_NonGI_MTD_Content_Report/index.html

Primary source set: - 2023/2026 Chapter 11 family - 2023/2026 Chapter 9.5 GI MTD chapters where the current report makes claims about MTD classification boundaries

Broader source lanes allowed when needed: - NJAC_2023 and NJAC_2026 for GI requirement / waiver / variance framing - Chapter 8 where maintenance-plan obligations are referenced - Chapter 5 only if a claim truly depends on the cited hydrologic method rather than on Chapter 11 itself

Key Changes

Source-validation posture

• Use 2D_GeneratePhase2DNonGIMTDContent.md as the content contract and 2D_RenderPhase2DNonGIMTDReport.md as the standalone HTML presentation contract.
• Treat the Chapter 11 family plus Chapter 9.5 as the primary authority for the report’s substance.
• Use broader docs/Sources only to support or narrow claims that the current report makes about GI requirement consequences, waiver/variance treatment, or related rule context.
• Keep the report in narrative technical-comparison prose, not audit-memo prose. Broader-corpus support should appear as normal attribution, not as rubric language.

Core factual corrections to make

• Recast the overview so it does not overstate 2026 as a wholesale reorganization or a newly formalized GI retrofit regime.
• Correct the quantity-standard shorthand in Section 1:
• remove the current “2-year and 100-year storms in most cases” summary unless directly supported in the cited source lane;
• keep quantity discussion tied to the actual waiver-or-variance and chapter-specific role of the BMPs.
• Correct MTD classification and performance:
• keep the verified split that Chapter 11.3 covers non-GI MTDs and Chapter 9.5 covers GI MTDs in both editions;
• keep certification-letter, NJCAT-verification, and TARP-reference treatment tied to what the chapters actually say;
• remove the invented standalone “NJCATS” process framing unless a real source in the corpus supports that exact label and process.
• Correct MTD sizing language:
• keep what Chapter 11.3 actually supports: Department-issued certification letter, published verification report, WQDS peak-flow sizing, MTFR logic, future-connections prohibition, and tested-configuration match;
• remove the current report’s universal Q_t = WQV / duration formula and universal 150% bypass trigger unless they are found in the source corpus as chapter-wide rules.
• Correct blue-roof treatment:
• keep blue roofs as quantity-only, waiver/variance BMPs;
• retain the 72-hour drain-time rule in both editions;
• remove the unsupported “new 24-hour minimum release period” change narrative.
• Correct subsurface gravel wetland treatment:
• keep 90% TSS and the chapter’s nitrogen / denitrification discussion where supported;
• remove the unsupported numeric “2023 40-50% TN to 2026 50-70% TN” update story;
• if the source supports 90% nitrogen-removal language, present it carefully as chapter-stated performance rather than as a new comparative update.
• Correct wet-pond treatment:
• remove the claim that 2026 newly moved wet ponds into Chapter 11 for the first time;
• recast it as a labeling/clarity change if that is what the 2023 and 2026 chapter titles and text actually support;
• remove the current 21-day residence-time shorthand unless it is directly supported by the Chapter 11 source text or a clearly relevant broader source.
• Correct sand-filter treatment:
• recognize that the repo now contains the 2026 Chapter 11.4 extracted markdown, so the current “2026 file missing” caution is stale and must be removed;
• compare 2023 and 2026 Chapter 11.4 directly and restate any claimed delta based on that actual comparison.

Narrative voice and HTML parity

• Keep the report cadence consistent with the prior restored reports: technical synthesis, not correction log.
• Remove audit-style phrasing such as “the current report was wrong,” “reviewed for this pass,” or similar source-audit narration except where a narrow qualification is truly necessary.
• Preserve the current report structure as a narrative report built around the five prompted sections, while keeping the existing top-of-report metadata and source-documents block.
• Update the standalone HTML only after markdown is locked.
• Preserve the current standalone layout, navigation, and visual model while bringing the touched prose into factual parity with the corrected markdown.
• Clean touched mojibake in both markdown and HTML.

Test Plan

• Confirm the corrected markdown still follows the intended Phase 2D section structure and reads like a narrative technical report.
• Spot-check the current high-risk claims against the actual source corpus:
• GI vs non-GI MTD split across Chapter 9.5 and Chapter 11.3
• 72-hour blue-roof drain rule in both editions
• absence or presence of any real 24-hour blue-roof release requirement
• MTD certification-letter / verification / MTFR sizing language
• whether any chapter-wide 150% bypass trigger exists
• subsurface gravel wetland TSS and nitrogen performance language
• wet-pond chapter placement and naming across 2023 vs 2026
• 2026 Chapter 11.4 sand-filter-with-underdrain content now present in the repo
• Verify that unsupported chapter-wide formulas, trigger thresholds, retrofit narratives, and “new 2026 regime” claims are removed or materially narrowed.
• Verify that broader-corpus-supported replacements are attributed naturally in prose rather than presented as rubric notes.
• Verify the standalone HTML matches the corrected markdown on all touched claims and that touched mojibake is removed.
• Run python scripts/build_green_guides.py and confirm the regenerated site/reports outputs reflect the corrected docs-side files.

Assumptions And Defaults

• This is a correction-and-voice-preservation pass, not a rewrite of the report’s scope.
• The Chapter 11 family and Chapter 9.5 remain the primary authority; broader docs/Sources support is allowed only to substantiate rule-context claims that Chapter 11 alone does not fully carry.
• The restored report voice used in the earlier report passes is the style baseline for this report as well.
• If a current claim cannot be clearly supported from the source corpus, it should be deleted or restated more narrowly rather than preserved.