The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides an instruction to execute a local bash script located at ${SKILL_DIR}/scripts/update-check.sh. This script is intended for manual update checks and requires the Bash tool to run.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user data to construct prompts for the HeyGen Video Agent.
Ingestion points: User-provided topic_or_script arguments, conversational Discovery inputs, and content analyzed from external assets/URLs.
Boundary markers: Employs structural delimiters such as 'CRITICAL ON-SCREEN TEXT' blocks and 'Style blocks' to separate components of the prompt.
Capability inventory: Uses the mcp__heygen__* toolset and the heygen CLI to perform video generation and asset management.
Sanitization: Relies on specific framing directives (e.g., 'This script is a concept and theme to convey') to guide the external AI, but lacks explicit sanitization or filtering of the user's input text.

heygen-video