The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands including pdflatex, bibtex, and git to compile LaTeX documents and manage repository states. These operations are performed on user-provided file paths and project directories, which are essential for the skill's primary functionality but require caution if the LaTeX compiler environment permits system-level interactions (e.g., via shell-escape).
[INDIRECT_PROMPT_INJECTION]: The structure command processes external .tex files that may contain untrusted content.
Ingestion points: The <tex_file> input parameter in the structure command (found in SKILL.md).
Boundary markers: None. The instructions do not define specific delimiters or security contexts for the content within the LaTeX files being analyzed.
Capability inventory: The skill can read/write files, create directories, and execute shell commands (pdflatex, git, bibtex, diff) as specified in SKILL.md and references/structure-rules.md.
Sanitization: Absent. The skill is instructed to preserve the source content during structural reorganization without explicit filtering for embedded instructions or malicious LaTeX macros.
[CREDENTIALS_UNSAFE]: The push command documentation in SKILL.md suggests using a URL format that includes plaintext tokens (e.g., https://<user>:<token>@github.com/...). Although these are placeholders, instructing the agent to construct URLs this way can lead to sensitive credentials being recorded in shell history, process lists, or Git configuration files (.git/config) within the execution environment.

latex-project-manager