PDF

PDF Skill

When To Use

• Read or review PDF content where layout and visuals matter.
• Create PDFs programmatically with reliable formatting.
• Fill and validate interactive PDF forms.
• Validate final rendering before delivery.

Tools + Contract Requirements

Immediately before the first create/edit authoring command, run mark_artifact_operation_started.mjs successfully exactly once using the command below. Do not run it for read-only work. For edits, replace create with edit; adjust the expected count and output format to match the requested outputs.

node container_tools/mark_artifact_operation_started.mjs --operation-kind create --expected-output-count 1 --output-format pdf

Workflow

1. Prefer visual review: render PDF pages to PNGs and inspect them.
   • Use pdftoppm from the bundled runtime or system Poppler when available.
   • If unavailable, install Poppler or ask the user to review the output locally.
2. Use reportlab to generate PDFs when creating new documents.
3. Use pdfplumber or pypdf for text extraction and quick checks; do not rely on text extraction for layout fidelity.
4. After each meaningful update, re-render pages and verify alignment, spacing, and legibility.

Fill And Validate AcroForms

Visual review alone is not a correctness check for a fillable PDF. A page /Widget annotation can render a value from its appearance stream while the canonical /AcroForm/Fields tree is missing or contains a stale value.

1. Keep the result interactive by default; set flatten=True only when the user explicitly requests a completed, static form. Preserve the source PDF, and do not flatten a signed PDF without an explicit workflow decision.
2. Inspect both representations before filling: enumerate fields from reader.get_fields() and /Widget annotations from every page's /Annots, following /Parent and /Kids. If a widget and a canonical field have the same name but are distinct objects with no /Parent relationship, do not call reattach_fields() blindly: it can create a second top-level field with the same name. Report the ambiguity or produce a static result.
3. Recover genuinely orphaned widgets, fill all pages, and write the result with pypdf:

from pypdf import PdfReader, PdfWriter
from pypdf.generic import NameObject

reader = PdfReader(input_pdf)
writer = PdfWriter()
writer.clone_document_from_reader(reader)

# Restores widgets that are missing from /AcroForm/Fields.
writer.reattach_fields()
fields = writer.get_fields() or {}
missing = set(expected_values) - set(fields)
if missing:
    raise ValueError(f"Form fields not found after repair: {sorted(missing)}")

values_to_write = dict(expected_values)
if flatten:
    # Paint every existing value before removing every widget.
    values_to_write = {
        name: field.get("/V", "/Off" if field.get("/FT") == "/Btn" else "")
        for name, field in fields.items()
    }
    values_to_write.update(expected_values)

writer.update_page_form_field_values(
    None, values_to_write, auto_regenerate=False, flatten=flatten
)

if flatten:
    # pypdf's flatten=True paints appearances but does not remove widgets.
    writer.remove_annotations(subtypes="/Widget")
    writer.root_object.pop(NameObject("/AcroForm"), None)

with open(output_pdf, "wb") as stream:
    writer.write(stream)

4. Reopen the written PDF before delivery. For an interactive result, require every expected field to be present in get_fields() with the expected /V, enumerate page widgets again, and confirm their effective /V (the widget value or inherited /Parent value) agrees. Confirm each updated widget has a non-empty /AP /N appearance and render the final pages to catch stale or clipped appearances. Do not rely on /NeedAppearances or a successful PNG render as proof that logical field data was updated.
5. For a flattened result, require zero /Widget annotations and no remaining /AcroForm field tree after reopening, then render the final pages. Keep an editable copy when the user may need to revise the form.

Temp And Output Conventions

• Use tmp/pdfs/ for intermediate files; delete them when done.
• Write final artifacts under output/pdf/ when working in this repo.
• Keep filenames stable and descriptive.

Dependencies

Prefer the Codex bundled workspace/runtime dependencies when available. The primary runtime is expected to include:

• Python packages: reportlab, pdfplumber, pypdf
• Rendering tools: pdftoppm and pdfinfo from Poppler

If a dependency is missing, install only what is needed.

Python packages:

uv pip install reportlab pdfplumber pypdf

If uv is unavailable:

python3 -m pip install reportlab pdfplumber pypdf

System tools for rendering:

# macOS (Homebrew)
brew install poppler

# Ubuntu/Debian
sudo apt-get install -y poppler-utils

If installation is not possible in this environment, tell the user which dependency is missing and how to install it locally.

Environment

No required environment variables.

Rendering Command

pdftoppm -png "$INPUT_PDF" "$OUTPUT_PREFIX"

Quality Expectations

• Maintain polished visual design: consistent typography, spacing, margins, and section hierarchy.
• Avoid rendering issues: clipped text, overlapping elements, broken tables, black squares, or unreadable glyphs.
• Charts, tables, and images must be sharp, aligned, and clearly labeled.
• Use ASCII hyphens only. Avoid U+2011 and other Unicode dashes.
• Citations and references must be human-readable; never leave tool tokens or placeholder strings.

Final Checks

• Do not deliver until the latest PNG inspection shows zero visual or formatting defects.
• Confirm headers, footers, page numbering, and section transitions look polished.
• Keep intermediate files organized or remove them after final approval.

Final response

Final response citations

Place :codex-file-citation{...} inline in prose, not in a trailing list. Use purpose="source" for Q&A/no-op and purpose="output" for create/edit.

• [HARD REQUIREMENT] Create/edit: cite each final PDF exactly once with a plain output citation. Summarize representative changes; do not cite every page or add a separate filename, path, or Markdown link. Example: Created :codex-file-citation{path="/abs/path/report.pdf" purpose="output"}, with the completed analysis and appendix.
• Q&A/no-op: do not edit or re-export. Inspect the complete relevant pages, preserve material headings, table/figure labels, footnotes, sources, and sample sizes, and cite each source PDF once with a plain source citation.

PDF citations currently support only plain file citations. Do not add artifact_kind, page_number, or other locators. Never cite rendered PNGs, scratch files, builders, or QA intermediates unless asked.

Final response suggested followups

• After successfully creating, editing, or analyzing an artifact, finish your final response with 3 useful next actions tailored to the artifact and the user's goal. Completing the requested work is not a reason to skip these actions: they help the user continue from the result. Do not perform the actions unless the user chooses one.
• Each action must be a distinct, concrete request the user could send next. Do not repeat completed work or invent filler. Omit the list if the user declines suggestions, the requested work fails, or no genuinely useful next action remains.
• Present each action as an unescaped Markdown list item using this exact syntax: - :codex-followup[Short action]{prompt="Complete request for that action"}.