The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data from Linear issue titles and descriptions to perform triage tasks.
Ingestion points: Linear issues (title, description, team, and customer context) are fetched via the Linear MCP integration.
Boundary markers: The setup instructions do not specify any delimiters or safety warnings to be included in the constructed triage prompt to prevent the agent from obeying instructions embedded within issue content.
Capability inventory: The resulting automation has the capability to post comments and modify issue metadata (labels, priority, assignee) within the Linear workspace.
Sanitization: There is no mention of sanitizing or escaping the issue content before it is interpolated into the triage prompt.
[COMMAND_EXECUTION]: The setup workflow instructs the agent to execute shell commands to register the automation with the platform backend.
Evidence: Step 4 in SKILL.md uses curl (or curl.exe on Windows) to perform a POST request to ${OPENHANDS_HOST}/api/automation/v1/preset/prompt.
Context: The command utilizes platform-provided environment variables (OPENHANDS_HOST, OPENHANDS_AUTOMATION_API_KEY) to interact with the vendor's own infrastructure, representing standard integration behavior for this skill's author.

linear-triage