The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text from LaTeX and Markdown files to calculate length statistics and generate summaries. Malicious content within these files could potentially affect the agent's logic when it processes the resulting reports.
Ingestion points: The scripts/check_length.py script reads source documents provided by the user through the --input parameter.
Boundary markers: The generated reports do not implement specific delimiters or safety instructions to prevent the agent from following directives found in the analyzed text.
Capability inventory: The skill has the ability to write files, execute the pdfinfo binary, and interact with the GitHub CLI.
Sanitization: While the script filters out LaTeX syntax, it does not specifically detect or sanitize natural language prompt injection attacks.
[COMMAND_EXECUTION]: The script scripts/check_length.py invokes the pdfinfo utility via subprocess.run to determine PDF page counts. This command execution is performed using a list of arguments and does not utilize a shell, which minimizes the risk of command injection.

nsfc-length-aligner