The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's orchestration logic utilizes internal Python scripts, including scripts/check_pipeline_integrity.py, scripts/ars_anchorize_draft.py, and scripts/ars_apply_revision_patch.py, to manage document anchors and ensure the integrity of text during multi-round revisions.
[COMMAND_EXECUTION]: The formatter_agent provides instructions for and utilizes system-level academic tools such as pandoc, tectonic, and xelatex to perform document conversions between Markdown, LaTeX, PDF, and DOCX formats.
[SAFE]: The visualization_agent is designed to generate runnable Python (matplotlib/seaborn) or R (ggplot2) code for creating publication-quality charts. This code is generated dynamically based on the paper's quantitative results and follows APA 7.0 standards.
[SAFE]: The skill incorporates an 'Anti-Leakage Protocol' (Knowledge Isolation) which mandates that the drafting agent must prioritize provided research materials over its internal knowledge to prevent the fabrication of methodologies or citations.
[SAFE]: The skill processes potentially untrusted external data, such as reviewer comments and user research materials. It mitigates indirect prompt injection risks by employing structured data extraction and explicit 'Generator-Evaluator' separation of model calls.

academic-paper