The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/figure_export.py invokes the pdffonts command-line utility via subprocess.run to verify that fonts are correctly embedded in generated PDF files. This is a legitimate requirement for academic submissions to avoid Type 3 font warnings. The implementation is secure, using an argument list instead of a shell string and including a 10-second timeout to prevent hangs.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and visualize experimental data from external files, which creates a surface for potential indirect prompt injection.
Ingestion points: The skill reads data from benchmark results in .writing/figures/data/.csv and .writing/figures/data/.jsonl as specified in the SKILL.md workflow.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish untrusted data content from operational instructions.
Capability inventory: The skill possesses capabilities for writing files to the local system (matplotlib.pyplot.savefig) and executing specific system commands.
Sanitization: The provided scripts do not perform explicit validation or sanitization of data strings ingested from files before they are used in plot labels or potentially analyzed by the agent.

scientific-visualization