The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes a dashboard architecture that ingests untrusted data from multiple external sources while maintaining command execution capabilities, creating a surface for indirect prompt injection.
Ingestion points: The dashboard monitors content from external sources including Mail (inbox and traffic), Merge queue (Pull Requests), and Issues (backlog/labels).
Boundary markers: The instructions do not define boundary markers or explicit warnings to ignore instructions embedded in the monitored data.
Capability inventory: The dashboard features a command palette used to execute gc system commands directly.
Sanitization: The skill does not mention any sanitization, filtering, or validation of the external content before it is displayed or processed.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute the gc binary with various subcommands (gc dashboard, gc start, gc dashboard serve). These are administrative commands used to manage the dashboard and API server.

gc-dashboard