The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download the Exa MCP server via npx to facilitate academic search. Additionally, the included NeurIPS 2025 template Makefile contains an upgrade target that fetches official style files from media.neurips.cc using curl.
[REMOTE_CODE_EXECUTION]: Instructions for the agent include executing the mcp-remote tool to connect to a hosted MCP endpoint (https://mcp.exa.ai/mcp) and running downloaded styles update scripts. These operations target well-known academic and developer services.
[COMMAND_EXECUTION]: The skill makes extensive use of the terminal and process toolsets. It provides patterns for process monitoring (ps aux), experiment execution (nohup python), and Git operations (git add/commit/push) to maintain experiment history.
[DYNAMIC_EXECUTION]: The pipeline includes several Python one-liner templates executed via python3 -c for validating bibliography entries, checking for duplicate labels in LaTeX files, and performing statistical significance tests.
[PERSISTENCE]: The skill utilizes the platform's native cronjob tool to schedule periodic status checks for long-running experiments, allowing the agent to maintain state and report progress across different user sessions.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting untrusted data from the web (paper abstracts, extraction results). It mitigates injection risks by enforcing a multi-step programmatic verification workflow for citations and recommending isolation through delegation to sub-agents for specific drafting tasks.

research-paper-writing