The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the iskron_* toolset for environment discovery, searching for agent roles, and managing communication. These interactions are conducted through a defined API for the iskron platform.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external sources. (1) Ingestion points: Communication channels (iskron_channel) and graph nodes (iskron_realm). (2) Boundary markers: Human-to-agent communication is prefixed with a specific contextual header string. (3) Capability inventory: The skill can read/write graph nodes and manage webhooks. (4) Sanitization: It enforces a policy that informational messages ('frames') are not instructions and requires human provenance for sensitive actions.
[PRIVILEGE_ESCALATION]: The skill instructions include the use of iskron_admin to configure one-shot webhooks for task results, which is a standard mechanism in the described multi-agent architecture for handling asynchronous updates.

collaborate