The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes a Makefile (templates/neurips2025/Makefile) that uses curl to fetch official LaTeX style files from media.neurips.cc during an upgrade process. This domain is the official repository for the NeurIPS conference, a well-known service in the machine learning community. Additionally, the skill instructions suggest adding a remote Model Context Protocol (MCP) server from exa.ai for paper search.
[INDIRECT_PROMPT_INJECTION]: The core workflow involves the agent reading and analyzing untrusted content from research repositories (README files, documentation, and source code). The instructions do not explicitly provide boundary markers or directions to disregard embedded commands in those files, creating a potential ingestion surface for indirect prompt injection.
Ingestion points: Research repository files such as README.md, experimental logs, and documentation (SKILL.md).
Boundary markers: Absent in the workflow instructions.
Capability inventory: Shell command execution (find, grep, ls) and network access via Python requests.
Sanitization: No specific sanitization or content filtering is prescribed.

ml-paper-writing