The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs automated document analysis and style checks by executing local shell commands. Evidence: SKILL.md and various files in the author_profile/ and writing_checklists/ directories instruct the agent to use tools like grep, pdfinfo, pdffonts, and file to scan LaTeX source files and compiled PDFs for structural consistency, broken references, and metadata.
[EXTERNAL_DOWNLOADS]: The skill includes a feature for generating non-data figures using external AI services. Evidence: figure_synthesis_guide.md outlines a workflow for creating image generation prompts and provides a Python script template for directly interacting with the Gemini API to produce architecture and concept diagrams.
[PROMPT_INJECTION]: The skill reads and processes user-controlled project files, which serves as a surface for indirect prompt injection. Ingestion points: The agent is instructed to read all .tex, .md, .log, and .pdf files in the working directory (e.g., sections/*.tex). Capability inventory: The skill can execute shell commands, write to the filesystem, and interact with external APIs. Sanitization: The instructions do not specify any validation or sanitization of the content read from these files.

paper-writing