The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a bundled Python script scripts/paper_case.py to handle complex PDF processing, workspace organization, and report scaffolding. The agent is instructed to execute this script via shell commands to perform tasks such as building a 'Paper Case' or generating back-end data previews.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from user-provided PDF files and supplementary research materials.
Ingestion points: Primary ingestion occurs when the skill reads raw text and visual elements from user-provided PDF documents via the scripts/paper_case.py tool.
Boundary markers: The skill instructions and methodologies (e.g., method.md, axioms.md) enforce rigorous grounding rules. Every extracted fact must be linked to a specific 'locator' (page number, bounding box, or text unit ID) to prevent hallucination and maintain provenance.
Capability inventory: The associated Python script performs file read/write operations within the local workspace and utilizes the PyMuPDF library for text and image extraction.
Sanitization: The script normalizes text blocks and enforces strict validation for identifiers like DOIs and IDs, which reduces the risk of malicious content affecting the extraction logic.
[SAFE]: External URL references in the references/ directory point exclusively to well-known and authoritative research reporting standard organizations, such as the EQUATOR Network, CONSORT, STROBE, and PRISMA. These are used as official sources for research methodology guidelines.

rw-paper-extractor