The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill directs users to download the application and related components from the project's official GitHub repository (github.com/getopenscreen/openscreen) and the Microsoft Store. It also facilitates the download of a machine-learning model for local audio transcription. The documentation encourages verifying Linux packages using SHA-512 checksums provided in the release manifest.
[COMMAND_EXECUTION]: Instructions include the use of standard system package managers (brew, winget, apt, dnf, pacman) for installation. It also details the usage of the openscreen CLI tool for automating recording, caption generation, and video export tasks.
[PRIVILEGE_ESCALATION]: The skill identifies necessary system-level permissions required for operation, such as sudo for Linux package installation, and 'Screen Recording' and 'Accessibility' permissions on macOS. It also notes that specific Linux functionality may require the user to be part of the input group.
[DYNAMIC_EXECUTION]: A Node.js script is provided as an example for programmatically editing the project's JSON-based .openscreen files. This script performs local file reads and writes to manipulate video metadata like aspect ratios and trim regions.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes external audio data for transcription and allows for user-defined annotations.
Ingestion points: The application captures screen content and microphone audio (processed by a local Whisper model) and reads .openscreen project files.
Boundary markers: No specific delimiters or instructions for the agent to ignore content within transcripts are defined.
Capability inventory: The skill can execute shell commands via the openscreen CLI and perform file system operations via Node.js.
Sanitization: The instructions do not specify sanitization methods for the transcribed text before it is embedded into video files or sent to optional AI editing providers.

openscreen