The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains multiple Python scripts in the scripts/ directory intended for local validation of the grant proposal content and the skill's own configuration.
[COMMAND_EXECUTION]: The script scripts/run_checks.py executes other local scripts using subprocess.run(). This is implemented using safe list-based arguments without shell execution, effectively preventing command injection.
[DATA_EXFILTRATION]: The skill instructions define a protocol for reporting bugs to the author's GitHub repository (huangwb8/bensz-bugs) using the gh tool. This is a documented developer feature triggered by specific user requests and targets a whitelisted domain with no access to sensitive system files.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it reads and processes user LaTeX files from the provided project directory.
Ingestion points: Reads existing .tex files and project data from the user-specified project_root.
Boundary markers: Absent. The instructions do not specify the use of delimiters or 'ignore' instructions for the content read from the file system.
Capability inventory: Ability to write to three specific LaTeX output files as defined in config.yaml and execute local Python validation scripts.
Sanitization: Input from user files is processed without explicit sanitization or validation of the text content.

nsfc-research-content-writer