The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a Bash block that executes a local Python script (tools/extract_paper_style.py) to analyze the structural layout of reference papers. This is an optional feature controlled by user arguments and is used for structural guidance rather than arbitrary code execution.
[DATA_EXFILTRATION]: To perform the automated review step, the skill transmits the generated paper outline to an external large language model (gpt-5.4) using the mcp__codex__codex tool. This is a core functional feature of the skill used to improve outline quality.
[SAFE]: The skill primarily interacts with project-local research documents (e.g., NARRATIVE_REPORT.md, experiment JSON files) to map claims to evidence. It follows standard practices for an AI-assisted academic writing workflow.

paper-plan