The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were identified during the analysis. The skill follows best practices for tool execution and file management.
[COMMAND_EXECUTION]: The skill executes local utilities such as LibreOffice (soffice), Tesseract (OCR), and macOS qlmanage via the subprocess.run module. These calls are implemented securely using list-based arguments, which mitigates the risk of shell injection. The tools are invoked to support core functionality: rendering slides and extracting text from images.
[EXTERNAL_DOWNLOADS]: All operations are conducted locally on the host system. The scripts do not contain any network-related code (e.g., requests, urllib, curl) for downloading remote content or exfiltrating data.
[PROMPT_INJECTION]: The skill provides structured instructions for an academic persona and strict formatting rules. It does not contain patterns intended to bypass AI safety filters, ignore prior instructions, or extract system-level prompts.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text from OOXML data and OCR results. While it lacks cryptographic or rigorous boundary markers for this content, it instructs the agent to treat OCR as evidence rather than absolute truth and focuses on grounding notes in visual evidence, which aligns with standard document processing practices.

ppt-speech-writer