winter-super-cli 2026.5.15

package/resources/local/codex/plugins/cache/openai-primary-runtime/documents/26.505.10851/skills/documents/tasks/style_lint_normalize.md

@@ -0,0 +1,70 @@
+# Task: Style lint + normalize (format consistency)
+
+## When to use
+Use this when the user asks for:
+- "Make formatting consistent"
+- "Apply our style guide"
+- "Remove random bold/Calibri/spacing changes"
+- "Why do headings look inconsistent?"
+
+This bundle provides:
+- `scripts/style_lint.py` — report likely inconsistencies
+- `scripts/style_normalize.py` — conservative cleanup (optional)
+
+## The reliable workflow
+1. **Render to PNGs** (baseline) and inspect a few problem areas.
+2. Run the **lint** to see what is causing drift.
+3. Apply **normalization** only if it matches the user’s intent.
+4. **Re-render and inspect all pages**.
+
+## 1) Lint
+```bash
+python scripts/style_lint.py input.docx --json /mnt/data/style_report.json
+```
+What to look for:
+- Lots of `run_direct_formatting`: common cause of “why is this one different”.
+- Multiple fonts/sizes in `Normal` body text.
+- “Heading-like” paragraphs that are not actually Heading styles.
+
+## 2) Normalize (conservative)
+`style_normalize.py` always clears **run-level** direct formatting overrides (bold/italic/underline/font/size/color) so styles drive appearance.
+
+### A) Default normalization (recommended starting point)
+```bash
+python scripts/style_normalize.py input.docx out_normalized.docx
+```
+
+
+> Tip: `style_normalize.py` also accepts `--out` as an alias:
+> ```bash
+> python scripts/style_normalize.py input.docx --out out_normalized.docx
+> ```
+
+### B) Also clear paragraph-level overrides (use sparingly)
+This can change layout. Use only when the user wants style-driven spacing/indents:
+```bash
+python scripts/style_normalize.py input.docx out_normalized.docx --clear_paragraph_format
+```
+
+### C) Enforce a simple heading spacing rule
+Useful when headings are visually inconsistent (space-after drift):
+```bash
+python scripts/style_normalize.py input.docx out_normalized.docx --enforce_heading_spacing
+```
+
+## Visual QA gate
+```bash
+python render_docx.py out_normalized.docx --output_dir /mnt/data/out_norm
+```
+Success criteria:
+- No clipped/overlapping text
+- Headings and body text are consistent
+- Tables remain aligned
+
+## Pitfalls / gotchas
+- **Clearing run overrides can remove intentional emphasis.** If the user wants to keep bold/italic emphasis, don’t normalize globally; instead normalize only certain styles/sections.
+- Some docs intentionally mix fonts (e.g., code blocks). Consider whitelisting styles rather than global clearing.
+
+## Deliverables
+- Deliver **only the final DOCX** requested by the user.
+- PNGs / optional PDFs are internal QA only unless explicitly requested.

package/resources/local/codex/plugins/cache/openai-primary-runtime/documents/26.505.10851/skills/documents/tasks/tables_spreadsheets.md

@@ -0,0 +1,36 @@
+# Task: Tables ↔ spreadsheets (import/export)
+
+## Goal
+Move tabular data between Excel and Word reliably, without hand-copying.
+
+## Import XLSX → DOCX table (simple)
+Use the helper to convert a sheet into a Word table:
+
+```bash
+python scripts/xlsx_to_docx_table.py /mnt/data/input.xlsx /mnt/data/table.docx --title "Table: Results"
+```
+
+What it preserves (best-effort)
+- cell values (data_only)
+- basic alignment (left/center/right)
+- header rows as bold
+- column widths (heuristic)
+
+What it does **not** preserve
+- merged cells, formulas, charts, conditional formatting, complex number formats
+
+## Export DOCX table → CSV
+
+```bash
+python scripts/docx_table_to_csv.py /mnt/data/input.docx --table_index 0 --out /mnt/data/table0.csv
+```
+
+## Render → PNG review checklist (tables)
+- Table fits within margins (no clipped columns)
+- Header row is visually distinct
+- Numbers align consistently (esp. decimals)
+- No unexpected wrapping that hurts readability
+
+## Common pitfalls
+- Word tables do not auto-match Excel column widths; you must verify visually.
+- Multi-line cells and merged cells round-trip poorly.

package/resources/local/codex/plugins/cache/openai-primary-runtime/documents/26.505.10851/skills/documents/tasks/templates_style_packs.md

@@ -0,0 +1,34 @@
+# Templates / Style Packs (DOTX) — apply consistent styles
+
+## Goal
+Apply a `.dotx` (or template `.docx`) style pack onto an existing report to improve professionalism and reduce bespoke styling.
+
+## Key idea
+A Word template is mostly:
+- `word/styles.xml` (style definitions)
+- `word/theme/theme1.xml` (colors/fonts)
+- optional: `word/fontTable.xml`, `word/numbering.xml`
+
+Direct formatting (manual bold/size/etc. on runs) can override style packs and cause inconsistent results.
+
+## Steps
+1. Apply the template parts to your doc:
+   ```bash
+   python scripts/apply_template_styles.py --template template.dotx --target report.docx --out report_styled.docx
+   ```
+
+2. Render and review:
+   ```bash
+   python render_docx.py report_styled.docx --output_dir out_styled
+   ```
+
+## Render → PNG review checklist
+- Typography looks consistent (headings/body)
+- Spacing and margins still acceptable
+- Tables didn’t reflow in a way that breaks readability
+- Page count changes are acceptable
+
+## Pitfalls
+- Style packs can change pagination. That’s expected.
+- If the report contains heavy direct formatting, consider normalizing by re-applying paragraph styles (Heading 1/2/3, Normal) before applying the template.
+- Custom style IDs in the target may be overwritten.

package/resources/local/codex/plugins/cache/openai-primary-runtime/documents/26.505.10851/skills/documents/tasks/toc_workflow.md

@@ -0,0 +1,61 @@
+# Task: Insert + update a Table of Contents (TOC)
+
+## Goal
+Create a TOC that **actually populates and stays correct** after edits.
+
+## Key reality
+A TOC is a **field**. It will not update unless fields are refreshed.
+
+## Headless-safe alternative (no Word field update)
+If you need a deterministic TOC in a fully automated / headless flow, prefer the **static TOC** workflow:
+
+```bash
+python scripts/internal_nav.py /mnt/data/input.docx --out /mnt/data/with_static_toc.docx
+```
+
+This builds a static TOC + internal links (TOC -> headings, headings -> Back to TOC) without relying on Word field updates.
+See: `tasks/navigation_internal_links.md`.
+
+## Requirements for a working TOC
+1. **Use Heading styles** (`Heading 1/2/3`) for headings. Do not fake headings with bold + bigger font.
+2. Keep heading text in the paragraph (avoid leading manual numbers as plain text).
+3. After edits, **update fields** before final export.
+
+## Insert a TOC at a placeholder
+1) Add a single paragraph containing the placeholder token:
+
+```
+[[TOC]]
+```
+
+2) Run the inserter:
+
+```bash
+python scripts/insert_toc.py /mnt/data/input.docx --out /mnt/data/with_toc.docx
+```
+
+Defaults: include Heading 1–3.
+
+3) Open in Word and update fields:
+- `Ctrl+A` → `F9` (Update Fields)
+- Save
+
+4) Render and visually verify:
+
+```bash
+python render_docx.py /mnt/data/with_toc.docx --output_dir /mnt/data/out
+```
+
+## Render → PNG review checklist (TOC)
+- TOC is present (not blank)
+- Indentation reflects heading levels
+- Page numbers in TOC match the actual headings’ pages
+- Headings that should appear do appear (and vice versa)
+- No placeholder text remains (e.g., “TOC will populate…”)
+
+## Common pitfalls
+- **Headings not styled** → TOC is empty.
+- **Manual numbering/direct formatting** → TOC levels/indentation drift.
+- **Fields not updated** → TOC and page numbers stale after edits.
+
+Tip: run `scripts/heading_audit.py` if you suspect heading-style issues.

package/resources/local/codex/plugins/cache/openai-primary-runtime/documents/26.505.10851/skills/documents/tasks/verify_render.md

@@ -0,0 +1,54 @@
+# Task: Verify / render a DOCX (DOCX → PNG)
+
+## Why this exists
+DOCX editing tools can "succeed" while the visual output is broken. Always verify by rendering.
+
+## Preferred: use the packaged renderer
+This uses a dedicated LibreOffice profile + writable HOME and produces `page-<N>.png` images:
+
+```bash
+python render_docx.py /mnt/data/input.docx --output_dir /mnt/data/out
+# For debugging LibreOffice failures:
+python render_docx.py /mnt/data/input.docx --output_dir /mnt/data/out --verbose
+# Optional: also write <input_stem>.pdf to --output_dir (for debugging/archival):
+python render_docx.py /mnt/data/input.docx --output_dir /mnt/data/out --emit_pdf
+```
+
+## Manual render command (if you need it)
+Use a unique LibreOffice profile (permission/locking issues are common in containers):
+
+```bash
+OUTDIR=/mnt/data/out
+INPUT=/mnt/data/input.docx
+BASENAME=$(basename "$INPUT" .docx)
+LO_PROFILE=/mnt/data/.lo_profile_${BASENAME}_$$
+mkdir -p "$OUTDIR" "$LO_PROFILE"
+
+HOME="$LO_PROFILE" soffice --headless -env:UserInstallation=file://"$LO_PROFILE" \
+  --convert-to pdf --outdir "$OUTDIR" "$INPUT"
+
+pdftoppm -png "$OUTDIR/$BASENAME.pdf" "$OUTDIR/$BASENAME"
+```
+
+## Success criteria
+- PNGs exist for each page
+- Spot-check page count and representative pages
+
+**Note:** LibreOffice sometimes prints scary-looking stderr (e.g., `error : Unknown IO error`) even when output is correct. Treat the conversion as successful if the PNGs exist and look correct (and if you used `--emit_pdf`, the PDF exists and is non-empty).
+
+## What to check in the PNGs
+- clipped text (especially headings and table cells)
+- overlapping objects
+- broken tables (wrapping, misalignment, missing borders)
+- unexpected font substitution
+- header/footer alignment and page breaks
+
+## Caveats
+- **Comments often don’t render** in headless LibreOffice PDFs. Use structural checks for comments.
+- **Field codes (page numbers, TOC)** may show placeholder values in some PDF renders. If the user needs proof, re-check in Word or update fields before final render.
+- **Multi-section docs** can have different page sizes/orientations; DPI is computed from the first section by default. If some pages look scaled oddly, use `--dpi` to override.
+
+## Delivery checklist
+- Final DOCX is clean (no internal citation tokens, no placeholder text)
+- Final render looks correct on all pages
+- `/mnt/data` contains only final outputs (unless user asked for intermediates)

package/resources/local/codex/plugins/cache/openai-primary-runtime/documents/26.505.10851/skills/documents/tasks/watermarks_background.md

@@ -0,0 +1,35 @@
+# Watermarks + Background Elements
+
+## Goal
+Detect and (carefully) remove watermark-like background elements from a DOCX, then verify the output via **render + diff**.
+
+## What watermarks are in Word
+Watermarks are often implemented as **VML shapes in headers**:
+`w:hdr → w:p → w:r → w:pict → v:shape → v:textpath string="DRAFT"`
+
+Other documents may use DrawingML shapes or background images.
+
+## Steps
+1. Audit the document:
+   ```bash
+   python scripts/watermark_audit_remove.py input.docx --mode report
+   ```
+
+2. Remove (heuristic) by matching a substring inside the watermark text:
+   ```bash
+   python scripts/watermark_audit_remove.py input.docx --mode remove --contains DRAFT --out cleaned.docx
+   ```
+
+3. Render and diff (recommended for QA/regressions):
+   ```bash
+   python scripts/render_and_diff.py input.docx cleaned.docx --outdir diff_watermark
+   ```
+
+## Render → PNG review checklist
+- If the watermark is visible in your renderer, confirm it is gone in the cleaned version.
+- Confirm headers/footers still render correctly (no missing logos/lines).
+- Confirm page count and layout remain acceptable.
+
+## Pitfalls
+- Removal is heuristic: it can delete legitimate header graphics if they match the substring.
+- Some VML watermarks won’t show in LibreOffice headless. When in doubt, validate in Word.

package/resources/local/codex/plugins/cache/openai-primary-runtime/documents/26.505.10851/skills/documents/troubleshooting/libreoffice_headless.md

@@ -0,0 +1,42 @@
+# Troubleshooting: LibreOffice headless rendering
+
+## Symptom: `soffice` hangs, times out, or errors in a container
+This is commonly caused by LibreOffice failing to create/lock its user profile, or attempting to write config/cache under a non-writable `HOME`.
+
+## Fix (recommended): use the packaged renderer script
+Use the canonical helper (`render_docx.py`). It:
+- creates a unique per-run LibreOffice profile
+- forces a writable `HOME` / XDG dirs under that profile
+- captures stdout/stderr so failures are diagnosable
+
+```bash
+python render_docx.py /mnt/data/input.docx --output_dir /mnt/data/out
+# If you're debugging a conversion failure:
+python render_docx.py /mnt/data/input.docx --output_dir /mnt/data/out --verbose
+```
+
+## Fix (manual): profile + writable HOME
+If you must run `soffice` directly, do this:
+
+```bash
+OUTDIR=/mnt/data/out
+INPUT=/mnt/data/input.docx
+BASENAME=$(basename "$INPUT" .docx)
+LO_PROFILE=/mnt/data/.lo_profile_${BASENAME}_$$
+mkdir -p "$OUTDIR" "$LO_PROFILE"
+
+HOME="$LO_PROFILE" soffice --headless -env:UserInstallation=file://"$LO_PROFILE" \
+  --convert-to pdf --outdir "$OUTDIR" "$INPUT"
+```
+
+## About scary stderr on "successful" conversions
+LibreOffice sometimes prints scary-looking messages (notably `error : Unknown IO error`) even when the output PDF is correct.
+
+Prefer these success criteria over stderr:
+- command completes
+- downstream PNGs exist and look correct
+
+## If you still get weird behavior
+- Ensure the profile directory is unique per process (use `$$` or a uuid)
+- Delete stale profiles between runs
+- Prefer `/mnt/data` over `/tmp` if you suspect permission sandboxing

package/resources/local/codex/plugins/cache/openai-primary-runtime/documents/26.505.10851/skills/documents/troubleshooting/run_splitting.md

@@ -0,0 +1,15 @@
+# Troubleshooting: run splitting ("why isn't my replace working?")
+
+## Reality
+Word splits text into runs unpredictably (style changes, proofing boundaries, fields, etc.).
+So searching for a substring and replacing it "as text" often fails.
+
+## Practical strategies
+- Work at the `<w:r>` / `<w:t>` level, not the paragraph text level.
+- When you must replace a token, consider inserting a hidden marker run first (during `python-docx` authoring) so you can reliably locate the target later when patching OOXML.
+- For tracked changes replacements, wrap **exact runs** you want deleted as `<w:del>`, then insert new `<w:ins>` adjacent.
+
+## Helper script
+`scripts/docx_ooxml_patch.py` contains utilities that:
+- find paragraphs by simple predicates (e.g., indentation)
+- replace the Nth tracked insertion inside a paragraph

package/resources/local/codex/plugins/cache/openai-primary-runtime/presentations/26.505.10851/.codex-plugin/plugin.json

@@ -0,0 +1,48 @@
+{
+  "name": "presentations",
+  "version": "26.505.10851",
+  "description": "Create, edit, render, verify, and export presentation slide decks. Use when Codex needs to build or modify a deck, slidedeck, presentation deck, slide deck, slides, PowerPoint, PPT, PPTX, .ppt, or .pptx file.",
+  "author": {
+    "name": "OpenAI",
+    "email": "support@openai.com",
+    "url": "https://openai.com/"
+  },
+  "homepage": "https://openai.com/",
+  "repository": "https://github.com/openai/openai",
+  "license": "MIT",
+  "keywords": [
+    "slides",
+    "slide",
+    "deck",
+    "slidedeck",
+    "presentation",
+    "presentations",
+    "powerpoint",
+    "ppt",
+    "pptx"
+  ],
+  "skills": "./skills/",
+  "interface": {
+    "displayName": "Presentations",
+    "shortDescription": "Create and edit presentations",
+    "longDescription": "Create, edit, inspect, render, verify, and export PPTX presentation slide decks locally. Use Presentations when the durable output or target is a deck, slidedeck, presentation deck, slide deck, slides, PowerPoint, presentation, speaker notes, or visual deck repair.",
+    "developerName": "OpenAI",
+    "category": "Productivity",
+    "capabilities": [
+      "Interactive",
+      "Write"
+    ],
+    "websiteURL": "https://openai.com/",
+    "privacyPolicyURL": "https://openai.com/policies/privacy-policy/",
+    "termsOfServiceURL": "https://openai.com/policies/terms-of-use/",
+    "composerIcon": "./assets/icon.png",
+    "logo": "./assets/logo.png",
+    "defaultPrompt": [
+      "Create a slide deck about the solar system",
+      "Research this company's latest earnings and create a slide deck on the company's financial performance",
+      "Create a visually stunning slide deck about the latest AI trends"
+    ],
+    "brandColor": "#C43E1C",
+    "screenshots": []
+  }
+}

package/resources/local/codex/plugins/cache/openai-primary-runtime/presentations/26.505.10851/README.md

@@ -0,0 +1,17 @@
+# Presentations
+
+Presentations is the file-type wrapper plugin for presentation slide deck workflows.
+
+This plugin is bundled into the Codex primary runtime.
+
+## Included Skills
+
+- `Presentations`: create, edit, render, verify, and export editable presentation slide decks.
+
+## Discoverability
+
+Use this plugin for presentation-oriented terms from the file-type naming model: slides, deck, PowerPoint, presentation, presentations, PPT, and `.pptx`.
+
+## Source And Copy Flow
+
+This plugin package is generated from `lib/artifacts/skills` and copied into `plugins/presentations` through `lib/artifacts/skills/mapping.json`.