The Agent Skills Directory

[COMMAND_EXECUTION]: The main script scripts/run.py utilizes subprocess.run to execute internal validation commands and pytest. This is a legitimate use case for self-testing and configuration validation within the skill's development environment.
[EXTERNAL_DOWNLOADS]: The skill includes a feature in scripts/core/reference_validator.py that can optionally verify DOIs by making HTTP requests to api.crossref.org. This is a well-known and trusted academic service used for metadata verification.
[DATA_EXFILTRATION]: No evidence of unauthorized data exfiltration was found. File system access is localized to the user's specified project directory. The skill implements a WritePolicy in scripts/core/security.py that strictly limits write operations to a single target file (extraTex/1.1.立项依据.tex) and explicitly forbids modification of core project files like main.tex or style files.
[PROMPT_INJECTION]: The skill uses structured prompts to interact with AI models for tasks like writing coaching and semantic diagnostics. These prompts are well-defined and do not contain instructions to bypass safety filters or ignore system-level constraints.

nsfc-justification-writer