The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to manage the slide environment, including using lsof to monitor network ports, mkdir for directory structure, and running locally generated JavaScript and Python scripts.
[EXTERNAL_DOWNLOADS]: The skill automates the installation of various third-party dependencies from official registries, including python-pptx, playwright, img2pdf, and the vercel CLI tool.
[DATA_EXFILTRATION]: In export mode, the skill provides a feature to deploy local slide content directly to Vercel. This involves uploading local files to a third-party cloud provider to generate a public sharing URL.
[REMOTE_CODE_EXECUTION]: The skill generates and executes scripts at runtime, specifically snapshot.mjs for browser automation via Playwright and extract_pptx.py for PowerPoint content extraction.
[PROMPT_INJECTION]: The ppt conversion mode creates a vulnerability surface for indirect prompt injection by ingesting untrusted data from external PowerPoint files. This data is then processed by the agent to generate slide content and speaker notes.
Ingestion points: External .pptx files processed via the python-pptx extraction script in references/ppt-conversion.md.
Boundary markers: None identified; extracted text is passed directly into the narrative expansion prompts.
Capability inventory: The skill has capabilities for file writing, shell command execution, and network deployment via Vercel.
Sanitization: No specific sanitization or filtering of the extracted PowerPoint text is mentioned before it is processed by the agent.

slide-deck