The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads research papers from arxiv.org and other user-provided URLs. It also integrates with the MinerU service (mineru.net) for structured PDF parsing. Additionally, it instructs the agent to install several well-known libraries such as pdfplumber, requests, and playwright.
[COMMAND_EXECUTION]: The workflow depends on executing local Python scripts (paper_ingest.py, render_slides.py) to manage document processing, layout validation, and rendering. It also uses the playwright CLI to automate browser-based slide generation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests raw text parsed from external papers to perform summarization and storyboard generation. This creates an attack surface where a maliciously crafted paper could attempt to influence the LLM's behavior during the distillation phase.
Ingestion points: Raw text and image metadata extracted from PDFs (managed in scripts/paper_ingest.py).
Boundary markers: The skill uses structured JSON schemas to constrain output, but lacks explicit instructional delimiters to isolate external text from the agent's core instructions.
Capability inventory: Subprocess execution, local file system writes, and network access for PDF retrieval and API calls.
Sanitization: scripts/render_slides.py properly escapes text content when building HTML templates to prevent script injection during the rendering process.

paper-explainer