mikeonbreeze/cc-crossbeam/adu-corrections-pdf
Claude Code and agent skill by mikeonbreeze/cc-crossbeam
- community
Checking that the SKILL.md file is still in the repository.
Formats a draft corrections letter (markdown) into a professional PDF. Single-purpose formatting sub-agent — no research. Receives markdown from the research agent, generates a styled PDF using the document-skills/pdf primitive, and returns a screenshot for QA. If the main agent finds issues in the screenshot, it will re-invoke this skill with fix instructions.
Install
npx vskill@latest install mikeonbreeze/cc-crossbeam/adu-corrections-pdfSKILL.md
ADU Corrections PDF Generator
Overview
Take a draft corrections letter in markdown and produce a professional PDF. This is a formatting-only sub-agent invoked by the research agent at the end of the pipeline.
Depends on: document-skills/pdf — the primitive PDF skill. Load it for the actual PDF generation tools (reportlab for Python, pdf-lib for JavaScript). This skill adds the domain-specific formatting on top.
What This Skill Does and Does NOT Do
| Does | Does NOT | |
|---|---|---|
| Apply city letterhead styling | Research codes or review plans | |
| Format sections, tables, confidence badges | Modify correction content | |
| Generate paginated PDF | Make judgment calls about findings | |
| Return screenshot for QA | Decide what goes in the letter |
Inputs
| Input | Format | Required | |
|---|---|---|---|
| Draft corrections markdown | .md file path | Yes | |
| City name | String | Yes | |
| Project address | String | Yes | |
| Project info | Object: applicant, designer, engineer, scope, dates | Yes | |
| Output path | File path for the .pdf | Yes | |
| Fix instructions | String (only on re-invocation after failed QA) | No |
Outputs
| Output | Format | |
|---|---|---|
| Corrections letter PDF | .pdf at specified output path | |
| QA screenshot | .png of page 1 — returned to calling agent |
Workflow
Step 1: Read Draft Markdown
Read the .md file. Validate expected structure: header block, notice, numbered sections, summary tables. Log warnings on unexpected structure but don't fail.
If fix_instructions are provided (re-invocation after failed QA), apply them before generating. Fix instructions might say: "Header is cut off — add more top margin" or "Table overflows page — reduce font size."
Step 2: Generate PDF
Use the `document-skills/pdf` primitive skill for the actual PDF generation. Two recommended approaches:
Approach A: reportlab (Python — best for structured layouts)
Use reportlab.platypus (SimpleDocTemplate, Paragraph, Table, Spacer, PageBreak) to build the letter programmatically. See document-skills/pdf/SKILL.md for reportlab patterns. This gives the most control over pagination, headers/footers, and table formatting.
Key reportlab elements for this letter:
SimpleDocTemplatewith letter pagesize and marginsParagraphwith custom styles for section headers, item titles, body text, code citationsTable+TableStylefor the review summary and sheet manifest tables- Custom
PageTemplatefor the city letterhead header and draft footer on every page
Approach B: markdown → HTML → PDF (faster to implement)
- Convert markdown to HTML (use
markednpm or Pythonmarkdownlibrary) - Inject into HTML template with
references/letter-styles.css - Render to PDF using puppeteer,
md-to-pdf, orwkhtmltopdf
See document-skills/pdf/reference.md for pypdfium2 rendering and pdf-lib JavaScript options.
Step 3: Screenshot for QA
Convert page 1 of the generated PDF to a PNG. Use document-skills/pdf tools:
Or command line:
Return the screenshot path to the calling agent. The calling agent handles the QA decision — this skill just provides the screenshot.
Letter Format Spec
Page Layout
- Size: US Letter (8.5" x 11")
- Margins: 1" top/bottom, 0.75" left/right
- Font: Arial/Helvetica — 10pt body, 12pt section headers, 14pt city name
- Line spacing: 1.3
Header (every page)
AI Notice Banner (page 1 only)
Light yellow box with amber border:
NOTICE: This is an AI-generated DRAFT corrections letter. Items flagged [VERIFY] require closer inspection. Items flagged [REVIEWER] require licensed professional assessment. A human plan checker must review and approve before issuance.
Confidence Badges (inline, color-coded)
- CONFIRM → green (quick check, HIGH confidence)
- VERIFY → amber (needs closer look, MEDIUM confidence)
- REVIEWER → red (requires licensed professional)
- COMPLETE → gray (human must fill in entirely)
Footer (every page)
Section Order
- Planning / Zoning
- Building
- Site / Civil
- Fire / Life Safety
- Structural [REVIEWER]
- Process Notes (Informational)
- Review Summary (tables)
Dependencies
| Skill | What It Provides | |
|---|---|---|
document-skills/pdf | PDF generation primitives — reportlab, pypdf, pdf-lib, pypdfium2, command-line tools |
References
| File | Contents | |
|---|---|---|
references/letter-styles.css | CSS stylesheet — only used if taking the HTML→PDF approach |
Important Notes
- Never modify correction content. Format exactly as received. Content issues are the research agent's problem.
- Every page must say DRAFT. Header, footer, and notice banner. This document cannot be mistaken for an official city issuance.
- Fail gracefully. If styling breaks, produce a plain-formatted PDF anyway. A plain PDF beats no PDF.
- Accept fix instructions. When re-invoked after a failed QA, the main agent will provide specific fix instructions. Apply them and regenerate.
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingserver/skills/adu-corrections-pdf/SKILL.md
Popularity
- Repository stars
- 172
- Repository forks
- 62
- 7-day trend score
- 26
Stars and forks are for the whole GitHub repository, not this skill alone. Refreshed 18 Mar 2026, 7 months ago.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/mikeonbreeze/cc-crossbeam/adu-corrections-pdf)