The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes grep commands to search for version identifiers within the user's project files and runs xelatex to compile documentation.
[EXTERNAL_DOWNLOADS]: Instructions are provided to install LaTeX dependencies through official and well-known package managers such as Homebrew (macOS) and APT (Linux).
[PRIVILEGE_ESCALATION]: The skill uses sudo to facilitate the installation of system-level LaTeX packages and updates via apt and tlmgr.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code from project directories to generate application materials, creating a vulnerability surface where embedded content could influence the generation process.
Ingestion points: Files with extensions like .js, .py, .ts, and .json in the user-specified project path are read by grep and the analysis script.
Boundary markers: The Python script uses filename separators (// ========== path/to/file.js ==========) when concatenating code for the source materials.
Capability inventory: The agent invokes local shell commands (grep), a provided Python analysis script, and document compilers (xelatex).
Sanitization: The analysis script includes escaping for LaTeX Verbatim blocks (escape_latex_verbatim) to ensure code snippets do not prematurely close LaTeX environments.

chinese-copyright-application