The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/extract-writing-styles.py utilizes subprocess.run to execute local Python scripts (document_converter.py and gemini_batch_process.py) located in the .claude/skills/ai-multimodal/scripts/ directory. While it employs a list-based argument structure that reduces shell injection risks, it establishes a functional dependency on these external scripts.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) when processing user-supplied files to extract writing styles.
Ingestion points: The extract-writing-styles.py script reads and analyzes content from documents and media files (.pdf, .docx, .jpg, .mp4, etc.) stored in the assets/writing-styles/ directory.
Boundary markers: Absent. The prompts hardcoded in the script for document and media analysis do not use delimiters or instructions to ignore commands potentially embedded within the file content.
Capability inventory: The skill can execute command-line operations via subprocess and perform filesystem reads.
Sanitization: No filtering or validation is performed on the extracted content before it is passed to the AI models for stylistic analysis.

copywriting