The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it reads and processes untrusted data from the user-specified research repository (e.g., README.md, source code) to determine the paper's contribution and content. * Ingestion points: Phase 1 reads repository documentation, configs, and logs via find, grep, and Read tools. * Boundary markers: No explicit delimiters or instructions are used to separate untrusted data from the drafting instructions. * Capability inventory: The agent utilizes Bash, Read, and Write tools throughout the 6-phase process. * Sanitization: No sanitization or filtering is performed on ingested data before it is incorporated into the drafting context.
[COMMAND_EXECUTION]: The skill uses the Bash tool to run shell commands for mapping repository structures, copying templates, and compiling the final LaTeX manuscript. (Evidence: phases/01-repo-understanding.md and phases/06-conference-formatting.md).
[EXTERNAL_DOWNLOADS]: The skill suggests using the tlmgr utility to download and install missing LaTeX packages from well-known CTAN mirrors during the assembly phase. (Evidence: phases/06-conference-formatting.md).

scholar-writing