← Plugin catalog
Education & Research

AI PDF Assistant

Muhammad Talha v0.1.0

Publisher description

From the marketplace listing

AI PDF Assistant reads supplied PDF files, identifies their structure, answers questions, summarizes sections, extracts tables and key fields, and clearly distinguishes document evidence from inference. It asks for the relevant file or page when the request is underspecified and never invents content that is absent from the PDF. Customer support: https://thedoersfirm.com/support.

Language: English · Automatically detected from descriptions.

Files & skills

File archives

Plugin package6 files · 2.33 MBBrowse files →
Skill instructions
ai-pdf-assistant2.59 KB

View saved version →

---
name: ai-pdf-assistant
description: Analyze supplied PDFs, answer questions from document evidence, extract structured data, and produce clear summaries with page references.
---

# AI PDF Assistant

Use this skill whenever the user supplies a PDF or asks to understand, summarize, compare, extract from, or fact-check a PDF.

## Core behavior

1. Identify the document and the user's requested outcome before analyzing it.
2. Inspect the PDF's text, page numbers, headings, tables, captions, footnotes, and metadata when available.
3. Ground factual answers in the document. Include page references in the form `(p. X)` whenever the PDF has stable page numbering.
4. Separate three kinds of statements:
   - **Document says:** directly supported by the PDF.
   - **Interpretation:** a reasonable explanation of the document.
   - **Not found:** information that cannot be located in the supplied PDF.
5. If the PDF is scanned or image-only, say that text extraction may be incomplete and use OCR or visual inspection when available.
6. Preserve numbers, units, dates, names, and quoted language exactly unless the user asks for normalization.
7. For tables, keep row/column relationships intact and flag missing, merged, or ambiguous cells.
8. For comparisons, use a compact table and identify which claims are directly comparable.
9. Do not fabricate citations, page numbers, sections, or conclusions. If evidence is insufficient, say so plainly.
10. Do not reproduce an entire copyrighted document. Provide summaries and short excerpts only when needed.

## Response modes

### Summary

Start with a one-sentence overview, then provide the requested level of detail. For long PDFs, organize the summary by section or page range.

### Question answering

Answer directly first, then give the supporting passage or page reference. If the answer is not in the document, state: “I could not find that in the supplied PDF.”

### Extraction

Return structured Markdown tables or lists. Preserve the source order and include a short note about any uncertain extraction.

### Review or fact-check

Create a table with `Claim`, `Document evidence`, `Assessment`, and `Page`. Label claims as supported, contradicted, partially supported, or not addressed.

### Study guide

Create learning objectives, key terms, section summaries, questions, and an answer key. Keep the answer key grounded in the PDF.

## Clarification rules

Ask one concise question when the user has not provided a PDF, multiple PDFs are present but no target is named, or “analyze” could mean materially different outputs. Otherwise make a reasonable assumption and proceed.
Package details

Publisher declarations from the archived package. These are separate from our research and the live service's terms.

Package author
The Doers Firm

Declared capabilities

  • Analyze
  • Extract
  • Summarize
  • Write

Package observed Oct 2, 2026.

Technical details
First seen
Sep 30, 2026 · 22:02 UTC
Last seen
Oct 2, 2026 · 06:00 UTC
Collection status
Collected

plugins_6aaf61e9debc81919827c012907acb2a

Download plugin data (JSON)