Smartform From Pdf logo

Smartform From Pdf

Community
babamba2
smartform-from-pdf

Build a SmartForm from a PDF of the printed form (or a screenshot) — windows, ruled grids as templates, texts in the fonts of existing SmartStyles, then fields, interface and item tables decided with the user — upload it to the DEV system, print it there and score the print against the source page.

Overview

Publisherbabamba2
Repositorysuperclaude-for-sap
Skill namesmartform-from-pdf
Stars
69
Forks
26
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by babamba2 on GitHub. Read the source before you install it.

Installation

Install the Smartform From Pdf AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/babamba2/superclaude-for-sap/tree/main/skills/smartform-from-pdf
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/babamba2/superclaude-for-sap.git /tmp/superclaude-for-sap
mkdir -p .claude/skills
cp -r /tmp/superclaude-for-sap/skills/smartform-from-pdf .claude/skills/smartform-from-pdf
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Smartform From Pdf in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Smartform From Pdf on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Smartform From Pdf is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

SC4SAP PDF / Screenshot → SmartForm

Reads one page of a printed form and writes a SmartForm that prints the same page: the ruled areas become templates, every text sits where the source has it, in the closest font an existing SmartStyle offers. Then the texts that are data become &fields&, item lines become a table over an internal table, and the interface is added. The form is uploaded to the DEV system, printed there (no spool) and the print is compared with the source — the score and every difference are reported.

<Response_Prefix>Every response triggered by this skill MUST begin with [Model: <main-model> · Dispatched: <sub-summary>] per ../../common/model-routing-rule.md § Response Prefix Convention. This skill dispatches no agents (Dispatched: none).</Response_Prefix>

<Team_Mode>No teamMode integration. Reading, planning and writing are scripted; the main thread decides fields and items.</Team_Mode>

<Use_When>

  • "make a SmartForm from this PDF", "PDF → SmartForm", "rebuild this form layout", "이 PDF로 스마트폼 만들어", "스크린샷 보고 스마트폼 생성", "このPDFからスマートフォーム"
  • A form layout exists only as a print (another system, a scan, a sample) and must become a SmartForm </Use_When>

<Do_Not_Use_When>

  • Changing an existing SmartForm in place — this skill creates a new form (a replace needs the user's explicit OK and an own Z/Y form)
  • SAPscript (SE71) or Adobe forms (SFP)
  • QA / production profiles — the upload is a repository write (DEV only; the MCP refuses elsewhere) </Do_Not_Use_When>

<Decisions_Fixed>

TopicDecision
InputPDF (exact positions, fonts, rules); a screenshot / scan is transcribed by the main thread into the same layout JSON (positions estimated — say so)
PagesOne page per run (the first, or --page N)
StylesExisting SmartStyles only — no style is created. Each text gets the closest paragraph (+ character format); sizes the styles lack are reported
GraphicsNot uploaded — each image area becomes a window %GRAPHICn (resolution from the image's pixels); it shows an existing SE78 graphic the user names (Step 5 graphics), else stays empty
ShadingFilled boxes shade the template cells they cover — a grey as black at (1 − grey) %, a colour as itself
TargetNew form Z*/Y* (ask; default Z<SOURCE>_COPY), package $TMP unless the user gives a package + transport
CheckThe generated form's own print (RenderSmartForm, invented sample values) compared with the source page — not Form Painter
Output<artifact-base>/smartforms/<FORM>/ — layout.json, spec.json, shape.json, <FORM>.xml, <FORM>.printed.pdf, compare.json
</Decisions_Fixed>

Step 1 — Target. Source file and target form name from ARGUMENTS; ask once for what is missing (form name, package / transport if not $TMP, output language). Uppercase the name; refuse names not starting with Z or Y.

Step 2 — Read the page.

node scripts/smartform/read-pdf.mjs "<source.pdf>" --page 1 --out "<out>/layout.json"

Report: page size, text runs, rules, images. A screenshot / scan: read the image and write layout.json yourself in the same shape (page.widthMm/heightMm, runs[] with x, y (baseline), w, text, pt, bold, font; lines[]; images[]) — scale pixels to mm from the page size (A4 unless the user says otherwise) and say that positions are estimates. Never guess pt from letter height: read where a few long lines start and end, take w from SAP's glyph widths (scripts/smartform/glyph-widths.json, key PDFUC|HELVE|B, 1/1000 em) and set pt = measured mm / width at 1 pt — sizes come out within 0.1 pt, and a centred paragraph is placed by w. Handwritten notes / highlights are the user's remarks, not the form: leave them out and ask what they mean in Step 5.

Step 3 — Style catalog. The SmartStyles the form may use: GetSmartFormContent(summary: true) of 1–3 forms of the same kind the user names (their styles — paragraphs and character formats) → <out>/styles.json (array of the styles entries). Ask for reference forms when the user names none; a catalog from earlier runs under <artifact-base>/smartforms/*/ may be reused.

Step 4 — Plan the layout.

node scripts/smartform/plan-form.mjs "<out>/layout.json" --styles "<out>/styles.json" --form <FORM> --out "<out>/spec.json"

The report (stderr) lists templates, windows, texts placed in cells, shaded cells, image windows and the font mapping (HELVE 9B → ZSTYLE/AB+C1 (HELVE 9B)); a mapping to another size or weight is a difference the print will show — tell the user, with the style that would be needed.

Step 5 — Data (with the user). Read spec.json and decide, then ask the user to confirm in one table:

  • which texts are values (document numbers, dates, names, amounts — anything that changes per document) → fields: [{ text, to: "&IS_HEADER-VBELN&" }]; labels stay fixed texts;
  • item lines are always a loop, never fixed texts — every itemCandidates entry of the Step 4 report (kind: items = numbers in the line; repeated = judge: tax lines, partner lists) and anything else that looks like items. No item DDIC type at hand → ask for one, do not fall back to texts:
    • over a ruled grid → items: [{ window, rows: [from, to], data: "IT_ITEMS", into: "GS_ITEM", cells: { "<col>": "&GS_ITEM-FIELD&" } }] — the first item row becomes a one-row template in a loop (fixed row height as in the source), the rows above / below stay templates, the window becomes MAIN;
    • over free texts → textItems: [{ header: [top, bottom], items: [top, bottom], count, footer: [top, bottom], data, into, cells: [{ text, x?, to }] }] (mm from the candidate; x tells two equal sample texts apart; to: "" leaves a cell empty) — header (column titles), loop and footer (totals, payment terms) go into one MAIN window, so the totals move down with the items;
    • a CURR / QUAN field whose DDIC reference is outside the form (EKPO-NETPR → EKKO-WAERS) needs refFields: [{ value: "GS_ITEM-NETPR", unit: "IS_HEAD-WAERS" }], else the print stops ("reference field unknown");
  • letterhead and footer blocks (company address, managing directors, bank / IBAN, register, tax numbers): ask — recommend SAP's way, a text module per company code (modules: [{ window, box: [x0, y0, x1, y1], module: "&GV_TM_FOOTER&", moduleStyle: true }], the name built in init: { input, output, code } from the company code — REF_NAME takes a whole field only; shape-form adds the guard: a name with no text module in STXFADM is cleared and the node prints only with a name, so a company code without a module leaves the block empty instead of stopping the print), else fixed texts. Print once with an empty / unknown company code too. A test module can be uploaded with CreateSmartForm (XML with HEADER-FORMTYPE T and one text node %TEXT1; its "is Text Module" generate error is expected); give the user the module content with tab positions and line spacing;
  • numbers (CURR / QUAN / DEC fields) print right-justified in their output length — in a cell the leading blanks wrap the value out of sight: give them (C) (&GS_ITEM-NETWR(C)&);
  • the interface: one structure for the header values, one table for the items — existing DDIC types (GetStructure / GetDataElement to check); propose names, never invent a DDIC object; globals for work areas (globals: [{ name: "GS_ITEM", type: "<line type>" }]);
  • graphics: which existing SE78 graphic each %GRAPHICn window shows (the user names it, or a reference form's graphics list) → graphics: [{ window: "%GRAPHIC1", name: "ZLOGO", btype: "BCOL" }];
  • conditional texts — a print shows one case only: ask which texts print only sometimes (a remark for credit notes, a line for one country; a reference form's conditions tell) → conditions: [{ text: "<text after fields>", cond: [{ op1: "IS_HEAD-CRDRF", cop: "EQ", op2: "'C'" }] }] (the text node's own condition; print both cases in Step 7). Write <out>/shape.json, then:
node scripts/smartform/shape-form.mjs "<out>/spec.json" "<out>/shape.json" --out "<out>/spec.json"

Step 6 — Write and upload.

node scripts/smartform/write-form.mjs "<out>/spec.json" --out "<out>/<FORM>.xml"

CreateSmartForm(form_name, xml_file, package_name?, transport_request?) — active: false with errors → fix the spec (the messages name the window / line type) and upload again with overwrite: true (own new form only).

Step 7 — Print and compare. RenderSmartForm(form_name, values) with invented sample values (the fixed texts of the source for a layout-only form: no values) → copy the PDF to <out>/<FORM>.printed.pdf, then

node scripts/smartform/compare-layout.mjs "<out>/layout.json" "<out>/<FORM>.printed.pdf" --out "<out>/compare.json"

Report the score and the differences (moved texts, other font sizes, missing / extra rules), and the unscored checks: shading (share of the source's shaded area printed in the same grey) and images (image areas printed in place). A layout-only form (before Step 5) should reach 95+; fields and tables change the texts, so compare the layout-only form first when the user wants proof of the layout.

<Data_Rules> The source page may carry real business data — it is read locally and never printed into the form: every value the user marks becomes a field. Sample values for the check print are invented (see /sc4sap:smartform-preview Data_Rules); do not read documents with GetTableContents / GetSqlQuery to fill them. </Data_Rules>

<Output_Format>

SmartForm ZSD_INVOICE_NEW created on DEV (package $TMP, active, FM /1BCDWB/SF00000123)
Source: invoice.pdf page 1 (A4 portrait) · 104 texts · 22 rules · 1 image
Layout: 3 templates, 41 windows · fonts: 7 mapped exactly, 1 to another size (COURIER 11 → 12, no style has 11)
Data: 18 fields (IS_HEADER: ZSDS_INV_HEAD), item table IT_ITEMS (ZSDT_INV_ITEM) rows 5–9
Print vs source: 98.3 (texts 104/104, rules 21/22)
Files: .sc4sap/work/DEV/smartforms/ZSD_INVOICE_NEW/

Reply in the user's language. </Output_Format>

<MCP_Tools_Used>

  • CreateSmartForm, RenderSmartForm, GetSmartFormContent (styles of reference forms)
  • GetStructure, GetDataElement, GetTable (interface types)
  • DeleteSmartForm (only for the skill's own test forms, on request) </MCP_Tools_Used>

<Related_Skills>

  • /sc4sap:smartform-preview — expected print of an existing form (the reverse direction) </Related_Skills>

Task: $ARGUMENTS

Frequently asked questions

What does the Smartform From Pdf AI skill do?

Build a SmartForm from a PDF of the printed form (or a screenshot) — windows, ruled grids as templates, texts in the fonts of existing SmartStyles, then fields, interface and item tables decided with the user — upload it to the DEV system, print it there and score the print against the source page.

Why use Smartform From Pdf on TypingMind?

Because you install it once and use it with any model. Smartform From Pdf is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Smartform From Pdf in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/babamba2/superclaude-for-sap/tree/main/skills/smartform-from-pdf. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Smartform From Pdf?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Smartform From Pdf?

As many as you like. As long as a model supports skills, you can use Smartform From Pdf with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Smartform From Pdf AI skill free?

Yes. It is published on GitHub by babamba2 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇