The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute standard command-line utilities including latexmk, pdfinfo, pdftoppm, and rg to compile research papers and validate their visual layout. These tools are used appropriately for the skill's stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from research paper manuscripts and external artifacts (CSV, JSON, logs), which constitutes a potential attack surface. 1. Ingestion points: The skill processes research paper text, LaTeX code segments, and numeric data files. 2. Boundary markers: The workflow includes explicit instructions to 'Inspect before editing', 'Locate the source data', and 'Verify the scoped source' before processing content. 3. Capability inventory: The skill involves modifying paper source files and executing shell commands for rendering and validation. 4. Sanitization: The instructions emphasize maintaining 'exact evidence', 'measurement integrity', and performing mandatory visual inspection of rendered outputs to detect discrepancies.
[SAFE]: The skill follows security best practices by requiring user authorization for updates and focusing on provenance and exact data preservation. No obfuscation, persistence, or data exfiltration patterns were detected.

paper-visual-craft