The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests and parses unstructured external data through multiple agents.\n
The revision_coach_agent is designed to parse reviewer comments from arbitrary formats including email text and PDF pastes, which allows untrusted data to enter the writing pipeline and potentially influence subsequent agent decisions.\n
The intake_agent performs 'Style Calibration' by learning from past writing samples provided by the user, which is a significant data ingestion surface.\n- [COMMAND_EXECUTION]: The skill utilizes external tools and dynamic code execution to produce its deliverables.\n
The visualization_agent generates Python (matplotlib/seaborn) or R (ggplot2) code based on the paper's results and executes it during the 'VLM Figure Verification' step to render images, creating a surface where generated code incorporating paper content is executed.\n
The formatter_agent generates command-line instructions for pandoc, tectonic, and xelatex to perform complex document conversions and compilation, operating on the final manuscript content.

academic-paper