The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection by ingesting and processing untrusted data from external sources.\n
Ingestion points: The skill includes workflows that crawl external product URLs using headless Chrome (/product-launch-video, /website-to-video) and fetch content from GitHub Pull Requests using the gh CLI (/pr-to-video).\n
Boundary markers: There are no specific instructions or delimiters provided in the routing logic to prevent the agent from following malicious instructions that might be embedded in the crawled HTML or PR descriptions.\n
Capability inventory: The skill possesses capabilities to execute shell commands, perform network operations via crawling, and write files to the project directory, which could be abused if the agent is manipulated.\n
Sanitization: The instructions do not define sanitization or validation logic for data retrieved from external sources before it is used to influence the video generation plan.\n- [EXTERNAL_DOWNLOADS]: Synchronizes and updates skill definitions from an external repository on GitHub to ensure the agent has the latest capabilities.\n
Evidence: The skill utilizes npx hyperframes skills update and npx hyperframes init to check for staleness and download the latest skill sets from the heygen-com/hyperframes repository.\n- [REMOTE_CODE_EXECUTION]: Provides instructions for the agent to install missing logic at runtime from an external registry.\n
Evidence: When a workflow is missing, the skill instructs the agent to run npx skills add heygen-com/hyperframes --skill <workflow-name> to dynamically fetch and install the required code.\n- [COMMAND_EXECUTION]: Utilizes system-level tools to interact with external services and the local development environment.\n
Evidence: Uses the gh CLI to read Pull Request data and npx to manage the HyperFrames ecosystem and dependencies.

hyperframes