OpenParable
Skip to content
← Back to skills

PDF Understanding

Documents

PDF Understanding is an Agent Skill for documents published in the B-M-Capital-Research/honeclaw repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Read and analyze user-uploaded PDFs, preferably using the system-extracted text fragments, and output summaries, key points, and risk notes

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/B-M-Capital-Research/honeclaw/blob/HEAD/skills/pdf_understanding/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/pdf-understanding, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

PDF Understanding Skill

When the message contains a 【PDF extraction text】 block, the system has already attempted to extract text from the PDF. Follow the flow below.

Standard Flow

  1. Identify the user's goal first: summary, Q&A, risk scan, financial metric extraction, investment judgment, and so on
  2. Prefer the content in 【PDF extraction text】 and answer with a structured conclusion:
    • Core conclusion (1-3 points)
    • Key evidence (important text points from the PDF)
    • Uncertain or missing items
  3. If extraction failed or the text is insufficient, which is common for scanned PDFs:
    • State clearly that the current PDF text is insufficient
    • Ask the user for OCR text, a key-page screenshot, or a specific page number
  4. If the user's question requires external fact checking, add tools:
    • web_search(query="...") for news, announcements, or industry context
    • data_fetch(...) for real-time market or company data

Output Requirements

  • Do not guess conclusions that are not present in the PDF
  • When citing evidence, stay close to the original wording and avoid inventing numbers
  • Keep the output concise and actionable for a chat setting