The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download the core transformation logic from an unverified GitHub repository (https://github.com/YuhangChen1/Paper2All.git), which is not associated with a trusted organization or well-known service.
[REMOTE_CODE_EXECUTION]: The installation and usage instructions require executing code from the external repository, including installing dependencies via a requirements file and running various Python scripts (pipeline_all.py, pipeline_light.py, generate_schematic.py).
[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to perform system-level operations, such as installing packages (apt-get install libreoffice poppler-utils), managing conda environments, and executing the transformation pipeline, which provides a large attack surface if the downloaded scripts are malicious.
[PROMPT_INJECTION]: The skill is highly vulnerable to indirect prompt injection as its primary function is to ingest and process untrusted external data (academic papers in LaTeX or PDF format) using LLM-powered scripts.
Ingestion points: Research papers (LaTeX/PDF) provided by the user in the --input-dir as seen in SKILL.md and usage_examples.md.
Boundary markers: There are no boundary markers or explicit instructions to the LLM to ignore potentially malicious commands embedded within the paper content.
Capability inventory: The skill allows significant control over the environment, including file system access (Read, Write, Edit) and shell command execution (Bash).
Sanitization: The documentation does not mention any sanitization or validation of the content extracted from external papers before it is used to drive the generation process.

paper-2-web