Milestone Alpha follow-up 2 Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

1Context

On 2026-10-19 Fadel M. Megahed printed the Step 6 report to PDF (replication_materials/findings/2026-10-18-report-print-before.pdf) and found it poor. His items: (a) each section starts a new page; (b) the request shows twice, the second time after "You:"; (c) the aliasing is one flat list, and the sentence "No main effect is aliased with a 2FI." is a bullet; (d) it is not clear how a run sheet with many factors or many runs prints; (e) the print needs a header and a footer with EXPERTA, its icon, the date and the version, if the HTML view stays as it is; (f) the citation needs a link to the app, in APA and in BibTeX form; (g) the report could say how the specification was made; (h) the report could say that his professorship pays for the LLM calls.

His answers on 2026-10-19: (g) yes, a short section; (h) an acknowledgment without an amount; (d) print up to a limit, with the limit set by the user, and a better HTML view of long run sheets that does not slow the page or harm the print; blank Response and Notes columns in the printed run sheet.

How the report works today: reportHtml() (lib/exports/report-html.ts) makes one HTML document with inline CSS and no script (decision C11). Step 6 shows it in a sandboxed frame (sandbox="allow-same-origin allow-modals", no scripts), "Report (.html)" downloads the same text, and "Print / PDF" prints the frame. The only print rules are h2 { break-before: page }, a repeated table header and rows that do not split.

Goal: The printed report reads as a short technical document: sections flow, the request shows once with its clarification rounds, the aliasing has structure, wide and long run sheets print in a controlled way with bench columns, every page has a running header and footer in Chromium, the citation comes in APA and BibTeX with the app link, and a new section records how the specification was made.

Architecture: The report stays one HTML file with inline CSS and no script. A new module lib/exports/report-css.ts holds the CSS and the measured print constants. Print-only parts (column blocks, bench columns, the row limit) are present in the HTML and shown only by @media print; the screen view gets a scroll box with a fixed header row. The card keeps the model ids and the prompt and schema versions of the translation, with defaults, so stored projects still read. A CSS-only toggle (radio inputs) switches the citation form, because no script runs in the frame.

Tech Stack: Next.js 16.3.8, React 19.3.0, TypeScript 5.9.3, zod 4.6.5, vitest 4.1.11, Playwright 1.63.0 (Chromium, page.pdf() for the print measurements). No new package.

Spec: replication_materials/plans/2026-10-14-ff-design-translator-design.md (version 6; Task 8 makes version 7), Sections 7.5, 8.1, 8.3 and 9.3; decision C11 (one HTML file, inline CSS, no script, no external resource).

2Decisions

#	Decision	By
R1	No forced page break. Sections flow; a heading stays with the text after it; a small table, a code block and an alias group do not split across pages.	FMM (a)
R2	The Request section shows the request text once. The turns after the first (the clarification rounds) follow under "Clarification", as "Question from EXPERTA" and "Answer".	FMM (b)
R3	Aliasing has two parts, "Main effects" and "Two-factor interactions". Each starts with its summary sentence as a paragraph, without "Example: ...", because the report lists every group. The groups are a numbered list in columns of at least 20rem.	FMM (c)
R4	No JavaScript table package. The report has no script (C11), and the frame does not run scripts. A paged or virtual table shows only some rows, so the print would lose rows unless the package had its own print mode, and the package would have to be inlined into each report file. The screen view uses CSS only: a scroll box (at most 75vh), a sticky header row, a sticky Run column, alternate row shading, and the rows in groups of 256, each group with `content-visibility: auto` so the browser skips the layout of groups out of view. Task 1 measures that the print still shows every group and that the view of 32,768 rows is not slower. If either fails, the groups stay and `content-visibility` goes.	Claude; FMM asked (d)
R5	Print row limit, set in step 6: "Rows of the run sheet in print", with "All rows" and "First 64", "First 256", "First 1,024" (only the values below the number of rows), each with its estimated pages. Default: all rows when the sheet has at most 256 rows, else the first 256. A limited print shows a note under the table. The HTML view always has every row.	FMM (d); default by Claude
R6	The run sheet prints on landscape pages (a named page). When a design has more factors than fit on one landscape page (K, measured in Task 1), the print uses column blocks: each block repeats Run and Std order. The header row repeats on each page. Blank Response and Notes columns are in print only, in the last block.	FMM (d), bench columns
R7	Running header and footer from `@page` margin boxes, with app texts only (no user text inside CSS): top left "EXPERTA · Design report", top right the date, bottom left "expertadoe.app · version {APP_VERSION}", bottom right "Page {n} of {N}". Chromium and Edge print them; other browsers print without them. The screen view does not change. The icon is in the title block of page 1 (inline SVG, no `url()`).	FMM (e)
R8	The PDF file name: during `print()`, the app sets the page title to "{project name}: EXPERTA design report" and puts the old title back after.	Claude
R9	Citation in APA 7 and BibTeX, both with `https://expertadoe.app`. In the report, a CSS-only toggle; the print shows both forms and no controls. The home page gets the same two forms with a toggle. The report may contain one link, to `https://expertadoe.app`; no other `href`.	FMM (f)
R10	New section "How the specification was made", after "Specification", with its own check box in step 6. The card keeps the model id of each voter and the mode, the prompt, schema and Jev-question versions and the time of the interpretation, with defaults, so stored and imported projects still read (`PROJECT_SCHEMA_VERSION` stays 1). A project without these values says "Not recorded".	FMM (g)
R11	One acknowledgment line, no amount: "The API and hosting costs of EXPERTA are funded by the Raymond E. Glos Professorship in Business at Miami University, held by Fadel M. Megahed." It covers hosting too, so it is in every report (heading "Acknowledgment" in the Versions and citation part) and on the home page under "Maintained by".	FMM (h), text approved 2026-10-19
R12	Other improvements: the title block replaces the empty first page; the design notation prints in the accent colour (today it is white on a background that browsers do not print); the word length pattern is one row; "Std" becomes "Std order"; numbers use tabular digits; the Versions part is a table plus the run sheet SHA-256, and it no longer repeats the run order.	Claude
R13	No change of `report_version` (the R report data do not change) and no app version change; the authors choose the version at release.	Claude

3Global Constraints

No package changes. Implementers do not run npm install (Dropbox).
Secrets and outside contact. Never read .env. Subagents make no network call to Vercel, Upstash, GitHub or an LLM provider.
C11 stays. The report is one HTML document: inline <style>, no <script>, no <link>, no @import, no url(, and no href except https://expertadoe.app. Every inserted text goes through escapeHtml(). No user text goes inside the CSS.
Same text everywhere. The frame, "Report (.html)" and the report in the zip stay byte-identical for the same contents and settings (report.spec.ts checks it).
Texts under review stay. Do not change content/glossary.json, data/guarantee-table.json, the R23 text or the references.
Fixed texts. Every new text is a fixed string, exported, with a unit test of its exact text; no advice words.
Wording. About 80% ASD-STE100, plain academic tone; labels exactly as this plan gives them.
Accessibility. The step 6 controls have visible labels; the axe sweep stays at zero violations; the citation toggle works with the keyboard.
Measure, do not guess. Every print constant (rows per page, the number K) comes from Task 1, is stored in config/print-layout.json with its source CSV, and a test recomputes it.
Tests. All in replication_materials/tests/. Test first. npm run check -- --fast after each task; npm run check after Tasks 4, 7 and 8.
Commits. One per task, with the attribution lines of the session.

4Review Focus

A factor name or a level with </style>, quotes or a very long word: the report escapes it, nothing reaches the CSS, and a long word wraps in its cell, not across the page. Test in Task 2.
A project stored before this plan (no model ids in its card) and a project filled in by hand: the new section says "Not recorded" or "filled in by hand", and nothing breaks. Test in Task 5 and Task 6.
A design with exactly K and K + 1 factors, and a sheet with exactly the limit and one row more: blocks and the note start at the right number. Test in Task 4.
A pending clarification (the last turn is a question without an answer): it shows with "No answer yet." Test in Task 3.
Print in a browser without margin boxes or named pages: the content still prints whole, in portrait, without the running header. Check in Task 7 (the CSS has no rule that hides content when those features are missing).

5Task 1: Print measurements

Files:

Create: lib/exports/report-css.ts (the print constants of R4 to R7 and the CSS text), replication_materials/tests/perf/print-layout.spec.ts, scripts/derive-print-layout.mjs, config/print-layout.json, replication_materials/tests/unit/print-layout.test.ts
Data: replication_materials/tests/perf/print-layout.csv

Interfaces:

Produces: PRINT = { bodyPt: 10, tablePt: 9, cellPadding: "0.15rem 0.4rem", marginMm: { top: 16, right: 14, bottom: 18, left: 14 }, chunkRows: 256 } in report-css.ts; config/print-layout.json with rows_per_page_landscape (Letter and A4, the smaller one is used), max_factor_columns_landscape (K), content_visibility_print_ok: boolean, view_ms_with and view_ms_without (32,768 rows). printLayout is imported from the JSON by Tasks 4 and 7.

Step 1: Write the probe spec. It builds synthetic run-sheet pages with the CSS of report-css.ts: factor names of 12 characters with a 3-character unit, levels of up to 6 characters. Measures, each 3 times: (1) page.pdf() page count for 512 rows, landscape, Letter and A4, with Run, Std order, 8 factors, Response and Notes; rows per page = floor(512 / full pages), using the smaller; (2) for 4 to 40 factors, the largest factor count whose table is not wider than the landscape content width in print emulation (with Response and Notes); (3) for 2,048 rows in groups of 256 with content-visibility: auto, whether the PDF has the same page count as without it (print shows every group); (4) the time from setContent to the first paint of a 32,768-row sheet in the scroll box, with and without content-visibility. One CSV row per measurement.
Step 2: Run FFD_PERF=1 npx playwright test print-layout.spec.ts and node scripts/derive-print-layout.mjs. The rules: rows per page = the smallest of the three runs; K = the smallest of the three runs; content_visibility_print_ok = all page counts equal.
Step 3: Write print-layout.test.ts: the JSON says "status": "measured", equals what the derive function gives from the CSV, and K is at least 4.
Step 4: Commit test: measure rows per page, the factor columns of a landscape page and the cost of content-visibility for the report print.

6Task 2: Report frame, title block, running header and footer, PDF name

Files:

Modify: lib/exports/report-html.ts, lib/exports/report-css.ts, components/steps/ReportStep.tsx
Test: replication_materials/tests/unit/report-html.test.ts, replication_materials/tests/e2e/report.spec.ts

Interfaces:

Consumes: PRINT (Task 1).
Produces: REPORT_TITLE = "Design report"; RUNNING_HEADER_LEFT = "EXPERTA · Design report"; runningFooterLeft(version: string): string = expertadoe.app · version ${version}; pdfTitle(projectName: string): string = ${projectName}: EXPERTA design report; cssString(text: string): string (CSS string literal; accepts only [0-9A-Za-z .:·-], else throws).

Step 1: Write the failing tests in report-html.test.ts: the CSS has no break-before: page; it has @page with the four margin boxes and the texts above, the date of now in @top-right, and counter(page) and counter(pages); the body starts with a title block (<h1>Design report</h1>, the project name, the notation line, the subtitle line) and an inline <svg with aria-hidden="true"; the .notation rule has no background (accent text colour and a border only); a project name with </style><script> appears only escaped and only outside <style>; cssString("a\"b") throws. In report.spec.ts: during the stubbed print() the page title is pdfTitle(name), and after it the title is back.
Step 2: Run them. Expected: fail.
Step 3: Implement. Move the CSS to report-css.ts; R1 rules (h2, h3, h4 { break-after: avoid }, table.small, pre, .keep { break-inside: avoid }, orphans: 3; widows: 3); @page { size: auto; margin: ... } with the margin boxes; title block; logo from the paths of public/brand/favicon.svg (copied as inline markup, 48 by 48 at 14 mm); in ReportStep.print(), set document.title before win.print() and restore it in a finally.
Step 4: Run npm run check -- --fast and npx playwright test report.spec.ts. Expected: pass.
Step 5: Commit feat: the report flows without page breaks, opens with a title block, and prints a running header and footer; the PDF gets the project name.

7Task 3: Request, design properties and aliasing

Files:

Modify: lib/exports/report-html.ts, lib/report/explain.ts
Test: replication_materials/tests/unit/report-html.test.ts, replication_materials/tests/unit/report/explain.test.ts (or the existing explain test file)

Interfaces:

Produces: aliasSentences(report, options?: { example: boolean }) and mainAliasSentences(report, options?: { example: boolean }); the default { example: true } keeps step 3 as it is. Texts: CLARIFICATION_HEADING = "Clarification", QUESTION_LABEL = "Question from EXPERTA", ANSWER_LABEL = "Answer", NO_ANSWER = "No answer yet.", MAIN_EFFECTS_HEADING = "Main effects", TWO_FACTOR_HEADING = "Two-factor interactions".

Step 1: Write the failing tests. Request: a request without clarification shows its text once and no "You:"; a request with one round shows the text, then <h3>Clarification</h3> and a <dl> with "Question from EXPERTA" and "Answer"; a pending last question shows NO_ANSWER; the first turn is left out only when it is the user turn with the request text. Aliasing: <h4>Main effects</h4> then <p>No main effect is aliased with a 2FI.</p> (a paragraph, no <li>); <h4>Two-factor interactions</h4> then <p>The 21 2FIs fall into 7 groups of 3.</p> and <ol class="chains"> with 7 items; a resolution III design lists its main-effect chains in an <ol> too; a full factorial has its sentence and no list. Properties: the word length pattern is one table with a header row of lengths and one row of counts; the run order shows once in the whole report.
Step 2: Run. Expected: fail.
Step 3: Implement the sections and the example option.
Step 4: Run npm run check -- --fast. Expected: pass.
Step 5: Commit feat: the report shows the request once with its clarification rounds, and the aliasing in two parts with numbered groups.

8Task 4: Run sheet on screen and in print

Files:

Modify: lib/exports/report-html.ts, lib/exports/report-css.ts, components/steps/ReportStep.tsx, lib/exports/archive.ts (it passes the same report options)
Test: replication_materials/tests/unit/report-html.test.ts, replication_materials/tests/e2e/report.spec.ts

Interfaces:

Consumes: printLayout (Task 1).
Produces: reportHtml(project, sections, now, options: ReportOptions = { printRows: null }) with interface ReportOptions { printRows: number | null } (null = all rows); printRowChoices(nrows: number): (number | null)[] = the values of [64, 256, 1024] below nrows, then null; defaultPrintRows(nrows: number): number | null = nrows <= 256 ? null : 256; estimatedPages(rows: number): number = Math.ceil(rows / printLayout.rows_per_page_landscape). Texts: PRINT_ROWS_LABEL = "Rows of the run sheet in print", printRowsOption(rows: number | null, nrows: number): string = All rows (${nrows}, about ${p} pages) or First ${rows} (about ${p} pages) (numbers with thousands separators, "page" for 1), printNote(m: number, n: number): string = Rows 1 to ${m} of ${n} are printed. The full run sheet is in the .html, .csv and .xlsx files., STD_HEADER = "Std order", RESPONSE_HEADER = "Response", NOTES_HEADER = "Notes", blockCaption(i: number, b: number, first: string, last: string): string = Run sheet, part ${i} of ${b}: ${first} to ${last}.

Step 1: Write the failing tests.
- Screen: the sheet is in <div class="rs-scroll">; the body rows are in <tbody> groups of PRINT.chunkRows (a 600-row sheet has 3); the CSS has the sticky header and the sticky first column for screen only.
- Print, narrow (K factors): one table; each row has two empty td.print-only cells; the header has Response and Notes as th.print-only.
- Print, wide (K + 1 factors): the screen table has class="screen-only", and a div.print-only holds ceil(factors / K) tables, each with Run and Std order, the captions from blockCaption, Response and Notes only in the last.
- Limit: with printRows: 256 and 257 rows, the rows after 256 have class="print-hidden" in the narrow table, the print blocks hold 256 rows, and printNote(256, 257) is under the sheet in a p.print-only; with printRows: null there is no note.
- printRowChoices(16) is [null]; printRowChoices(4096) is [64, 256, 1024, null]; defaultPrintRows(256) is null and defaultPrintRows(257) is 256.
- e2e: step 6 shows the select with PRINT_ROWS_LABEL only when the sheet has more than 64 rows; its default follows defaultPrintRows; the download still equals the frame; the zip report equals the download for the chosen value.
Step 2: Run. Expected: fail.
Step 3: Implement. The named page for the run sheet section (@page runsheet { size: landscape }, section.run-sheet { page: runsheet }); content-visibility: auto on the groups only if printLayout.content_visibility_print_ok (Task 1). The step 6 select stores its value in the component state, next to the contents choice.
Step 4: Run npm run check. Expected: pass.
Step 5: Commit feat: long run sheets scroll in a box on screen; in print they use landscape pages, column blocks, bench columns and a row limit chosen in step 6.

9Task 5: Keep the record of the interpretation in the card

Files:

Modify: lib/card/card.ts (VotersSchema, votersOf, cardFromTranslation), lib/projects/schema.ts (comment only), spec 8.1
Test: replication_materials/tests/unit/card/card.test.ts (or the existing card test file), replication_materials/tests/unit/projects/schema.test.ts (or the existing one)

Interfaces:

Produces: VotersSchema lines get model_id: z.string().nullable().default(null); VotersSchema gets meta: z.object({ mode: z.enum(["single", "ensemble"]), prompt_version: z.string(), schema_version: z.string(), jev_questions_version: z.string(), interpreted_at: z.string().regex(LOCAL_TIME_PATTERN) }).nullable().default(null). cardFromTranslation(answer, now, newId) fills both (the time from isoWithOffset(now)).

Step 1: Write the failing tests. A card from a recorded fixture keeps each voter's model_id and the four versions and the time; a card stored before this change (no model_id, no meta) parses with null in both; a project file exported before the change imports with the same result; a card filled by hand still has voters: null.
Step 2: Run. Expected: fail.
Step 3: Implement. Add a sentence to spec 8.1 on the stored record of the interpretation.
Step 4: Run npm run check -- --fast. Expected: pass.
Step 5: Commit feat: the card keeps the model ids, the mode, the prompt and schema versions and the time of the interpretation.

10Task 6: Section "How the specification was made"

Files:

Modify: lib/exports/report-html.ts (REPORT_SECTIONS, SECTION_TITLES, CONTENTS_LABELS, a builder), components/spec-card/VoterSummary.tsx (its texts move to a pure function), lib/card/summary.ts (new)
Test: replication_materials/tests/unit/report-html.test.ts, replication_materials/tests/unit/card/summary.test.ts (new), replication_materials/tests/e2e/report.spec.ts

Interfaces:

Consumes: the card fields of Task 5.
Produces: voterSummary(voters: CardVoters): { heading: string; lines: string[]; jev: string | null } in lib/card/summary.ts, with the exact texts that step 2 shows today (VoterSummary renders from it); fieldLabel(path: FieldPath | "nfactors", card: Card): string with "Number of runs", "Minimum resolution", "Replications", "Repeat only", "Center points", "Randomized", "Seed", "Number of factors", and for a factor field ${factor name}: ${"name" | "low" | "high" | "unit" | "type"}. Section key provenance, title and contents label "How the specification was made". Texts:
- PROVENANCE_HAND = "The specification was filled in by hand. No language model was used."
- PROVENANCE_IMPORTED = "The specification comes from an imported project file without a record of its interpretation."
- provenanceIntro(date: string, mode: "single" | "ensemble"): string = EXPERTA interpreted the request on ${date} in ${mode === "ensemble" ? "ensemble" : "single-model"} mode. The user checked the specification in step 2 and confirmed it.
- NOT_RECORDED = "Not recorded"
- disagreeText(labels: string[]): string = Fields where the voters gave different values: ${labels.join(", ")}. The user chose the value of each.
- checkText(labels: string[]): string = Fields with a quote that did not pass the evidence check: ${labels.join(", ")}.
- editedText(labels: string[]): string = Fields that the user changed in step 2: ${labels.join(", ")}.
- NO_CHANGES = "The voters agreed on every field, and the user changed no field."
- versionsText(meta): string = Prompt version ${p}, schema version ${s}, Jev questions version ${j}.; without meta: VERSIONS_NOT_RECORDED = "The prompt and schema versions were not recorded for this project."

Step 1: Write the failing tests. A translated project: the section has the intro, a "Voters" table (Voter, Model, Result) with NOT_RECORDED for a missing model id, the step 2 heading and the Jev text from voterSummary, the disagree, check and edited lines from the card (or NO_CHANGES) and the versions line. A hand-filled project: PROVENANCE_HAND. An imported card: PROVENANCE_IMPORTED. Step 2 still shows the same texts (its existing tests pass unchanged).
Step 2: Run. Expected: fail.
Step 3: Implement.
Step 4: Run npm run check -- --fast. Expected: pass.
Step 5: Commit feat: the report records how the specification was made.

11Task 7: Citation in APA and BibTeX, the Versions part and the acknowledgment

Files:

Modify: lib/site.ts, lib/exports/report-html.ts, lib/exports/report-css.ts, app/page.tsx, components/site/CitationToggle.tsx (new client component)
Test: replication_materials/tests/unit/site.test.ts (new or existing), replication_materials/tests/unit/report-html.test.ts, replication_materials/tests/e2e/home.spec.ts (or the existing home page spec)

Interfaces:

Produces in lib/site.ts: APP_URL = "https://expertadoe.app"; CITATION_APA = \Vazquez, A. R., & Megahed, F. M. (2026). ${APP_FULL_NAME} (Version ${APP_VERSION}) [Computer software]. ${APP_URL}`(the title in italics in HTML);CITATION_BIBTEX` =
```
@software{vazquez2026experta,
  author  = {Vazquez, Alan R. and Megahed, Fadel M.},
  title   = {{EXPERTA}: Fractional Factorial Design Translator},
  year    = {2026},
  version = {<APP_VERSION>},
  url     = {https://expertadoe.app}
}
```
CITATION is removed; its users take CITATION_APA. Toggle labels "APA" and "BibTeX". ACKNOWLEDGMENT_HEADING = "Acknowledgment" and FUNDING_LINE = "The API and hosting costs of EXPERTA are funded by the Raymond E. Glos Professorship in Business at Miami University, held by Fadel M. Megahed.": in every report after the citation, and on the home page under "Maintained by". Versions part: a table "Software" (Software, Version) with the rows EXPERTA, R, WebR, FrF2, DoE.base, Report format, then Run sheet SHA-256: ${hash} when the project has it.

Step 1: Write the failing tests. The exact strings above; the report has two radio inputs and labels in the citation, the APA text with one <a href="https://expertadoe.app">, the BibTeX in <pre>, CSS that shows one form on screen and both in print with the controls hidden; the only href in the report is the app URL; the Versions table rows; no run order line in that part; <h3>Acknowledgment</h3> and FUNDING_LINE in every report (hand-filled too). Home page: FUNDING_LINE under "Maintained by"; the two buttons with aria-pressed, APA shown first, BibTeX after a click or Enter.
Step 2: Run. Expected: fail.
Step 3: Implement.
Step 4: Run npm run check. Expected: pass (axe sweep included). Review Focus 5: a unit test asserts that the print CSS hides only .screen-only, .print-hidden and the citation controls, and that the column blocks show through @media print alone (not through the named page or a margin box).
Step 5: Commit feat: the citation in APA and BibTeX with the app link, on the home page and in the report; a Versions table with the run sheet SHA-256.

12Task 8: Documents, reference prints and the record

Files:

Modify: spec Sections 7.5, 8.3 and 9.3 (version 7), README.md, lib/review/known-issues.ts and its tests (the print item goes), replication_materials/process/process.json, the ledger of this plan, the spec page files
Create: replication_materials/findings/2026-10-19-report-print-after-7f16.pdf, ...-after-32f64.pdf, ...-after-res3.pdf, ...-after-4096.pdf (from page.pdf() on reference projects: 7 factors in 16 runs; 32 factors in 64 runs; a resolution III design; 4,096 runs with the default limit)

Step 1: Spec: 8.3 describes the report sections (with "How the specification was made"), the print rules R1 and R4 to R8, and the row limit; 9.3 step 6 adds the select; 7.5 notes the clarification layout. Version 7 in the header and the change list.
Step 2: Remove the print item from KNOWN_ISSUES and its tests (the note keeps the other two items). This change ships in the same deploy as the redesign.
Step 3: Make the four reference PDFs with a small spec in replication_materials/tests/perf/ (run with FFD_PERF=1), and record their page counts in the ledger.
Step 4: Process event, ledger, spec page rebuild; npm run check; node scripts/secret-scan.mjs.
Step 5: Commit docs: the report print in the spec (version 7) and the README, reference prints, and the known issues without the print item.

13Execution notes

The author approved both plans and the order on 2026-10-19, and the acknowledgment text in R11.
Order: this plan runs after the plan of the deferred minors (2026-10-19-alpha-minors-plan.md), on its own branch, with its own pull request and deploy.
After the deploy: site.spec with FFD_EXPECT_CSP=1 against https://expertadoe.app, and one print of a real project by the author.