The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process outputs from various sub-agents and external entities, which creates an indirect prompt injection attack surface.
Ingestion points: Processes data and triggers from other agents via platform tools like iskron_channel and through the observation of 'frames'.
Boundary markers: The instructions specifically mandate that messages from external owners be passed to the user 'literally, with provenance, without silencing anything' to ensure transparency and prevent deceptive filtering.
Capability inventory: The agent has the ability to manage task states via iskron_case and can initiate new agent processes using specific command strings.
Sanitization: The skill relies on policy-based handling (literal transmission) rather than explicit data sanitization routines to mitigate risks from untrusted agent outputs.
[COMMAND_EXECUTION]: The skill involves the generation and execution of commands to manage the lifecycle of sub-agents (e.g., the start command for launching agent sessions). These operations are standard functional requirements for the skill's role as an orchestrator within its specific platform environment.

chief-of-staff