The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/export_pdf.py utilizes the Python subprocess module to call external binaries like pandoc and xelatex for document conversion. The script follows security best practices by passing command arguments as a list rather than a shell string, effectively mitigating shell injection risks. Similarly, scripts/validate_structure.py executes fc-list to verify system font availability.
[DATA_EXPOSURE]: The provided utility scripts are designed to read local Markdown files (e.g., article.md) for validation and processing. This is a legitimate functional requirement for a document creation skill and does not involve access to sensitive system directories or credential files.
[PROMPT_INJECTION]: The instructions in SKILL.md and the references/ folder define a specific persona and strict stylistic rules (such as zero-tolerance for specific 'forbidden' words). These instructions are scoped to the creative writing task and do not attempt to override agent safety protocols or bypass underlying system constraints.
[INDIRECT_PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests untrusted Markdown data for processing through its validation and export scripts. However, the capabilities are limited to local document processing, and the scripts do not facilitate network communication or arbitrary code execution from the processed content.

luozhenyu-bookwriter