The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the agent-capture-pp-cli binary using npx -y @mvanhorn/printing-press-library and go install github.com/mvanhorn/printing-press-library/.... These resources are hosted on official repositories associated with the author.
[DATA_EXFILTRATION]: The skill provides comprehensive functionality for screen capture, video recording of windows, and OCR (Optical Character Recognition). This creates a significant surface for exposing sensitive data displayed on the user's desktop to the agent.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Commands like ocr, screenshot, and watch ingest unstructured data directly from the graphical user interface.
Boundary markers: There are no instructions or delimiters provided to help the agent distinguish between intended UI text and potentially malicious hidden instructions in the captured content.
Capability inventory: The skill allows for command execution, file system access for saving media, and screen recording.
Sanitization: The skill does not mention any sanitization or filtering of text extracted via OCR before it is processed by the agent.
[DYNAMIC_EXECUTION]: The vhs command executes terminal automation scripts (tape files), and the remotion command renders video from React compositions. Both commands involve the execution of script-based templates to generate output.
[COMMAND_EXECUTION]: The skill instructs the agent to invoke a local CLI binary with various arguments. The inclusion of an --agent flag enables non-interactive mode (--no-input, --yes), which can lead to actions being performed without explicit user confirmation.

pp-agent-capture