The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/latex_checker.py uses subprocess.run to programmatically call scripts/clean_latex.py when the --fix flag is enabled. Although it avoids using shell=True and uses a list for arguments, it invokes external script execution based on the user-provided .tex file path.
[DATA_EXFILTRATION]: The load_tex function in scripts/latex_checker.py implements a recursive file resolution mechanism for LaTeX \input{} directives. This implementation lacks path validation or sanitization, making it vulnerable to directory traversal attacks. An attacker could provide a malicious LaTeX file containing \input{../../.ssh/id_rsa} to trick the agent into reading and potentially exposing sensitive file contents through the checker's output or error messages.
Ingestion points: The load_tex function in scripts/latex_checker.py (line 45) reads file paths from LaTeX commands.
Boundary markers: Absent; the content of included files is merged directly into the processing buffer without delimiters.
Capability inventory: subprocess.run (line 254) in scripts/latex_checker.py and file-write operations in scripts/clean_latex.py (line 213).
Sanitization: Absent; the script does not check if the resolved path for \input remains within the project directory.

latex-formatting