The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/style_extractor.py script uses the subprocess.run function to execute the pdftoppm utility, which is a well-known system tool for converting PDF documents into images for visual analysis.
[DATA_EXFILTRATION]: The test script scripts/test_super_ppt.py includes a hardcoded absolute file path to a document on the author's local filesystem. This reveals information about the vendor's internal directory structure and environment.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through its style extraction and AI prompt generation feature.
Ingestion points: Document metadata and text content are extracted from user-provided .pptx and .pdf files via the extract_style function in scripts/style_extractor.py.
Boundary markers: The generated AI style prompt lacks delimiters or instructions to ignore or isolate potentially malicious instructions embedded within the extracted text content.
Capability inventory: The skill includes utilities for filesystem modification (ppt_editor.py) and external command execution (style_extractor.py).
Sanitization: Extracted document content is directly interpolated into the AI-facing style prompt without validation or sanitization, creating a surface where malicious instructions in a document could influence the agent's behavior.

super-ppt