The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted research proposals to generate structured budget justifications. Ingestion points: User-provided research materials and proposal files. Boundary markers: No explicit delimiters or boundary instructions are used when the agent processes input materials. Capability inventory: The skill has the capability to write files to the local system and execute shell commands (xelatex). Sanitization: It implements a sanitize_paragraph_latex function to escape LaTeX special characters and restricts allowed commands to prevent malicious macro execution (e.g., via \write18).
[COMMAND_EXECUTION]: The skill uses subprocess.run to execute the xelatex binary for document compilation. Security is managed through strict validation of output directories and input file paths. The skill also provides instructions for the agent to use the GitHub CLI (gh) and a local utility bensz-collect-bugs to report software issues to a specific developer repository.

nsfc-budget