The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses curl to interact with an automation backend. The host URL and session API key are retrieved from the environment's runtime services (OPENHANDS_HOST and OPENHANDS_AUTOMATION_API_KEY), which is a secure and standard practice for internal service communication.
[COMMAND_EXECUTION]: The skill uses a curl command to register a prompt preset with a local automation service. This command is structured to use environment variables for authentication and target URL, avoiding hardcoded secrets.
[DATA_EXFILTRATION]: No evidence of unauthorized data exfiltration. The skill processes Linear issue data (title, description, etc.) specifically for the purpose of triage within the configured environment.
[CREDENTIALS_UNSAFE]: The skill does not contain hardcoded credentials. It correctly references environment variables ($OPENHANDS_AUTOMATION_API_KEY) for authentication with the automation service.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Linear issues. While this presents an indirect prompt injection surface, the skill is designed to suggest changes or post comments for human approval, which serves as a mitigation. No dangerous capabilities like eval or direct shell execution on issue content are present.

linear-triage