The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data, including repository origins, platform readiness states, and artifact links, to determine the flow of the onboarding process.
Ingestion points: Data is ingested via iskron_me(action="readiness") and iskron_orient, which includes user-controlled or environment-controlled attributes like repository origins and role descriptions.
Boundary markers: The skill contains explicit instructions to verify intent with the user before performing any write operations ("спроси до записи") and warns not to trust arbitrary markers or attributes to prove understanding.
Capability inventory: The agent has the ability to create graphs (iskron_realm), add project roles (iskron_add_karta), and manage organization members (iskron_admin).
Sanitization: The instructions require the agent to re-read state from the API after every action rather than relying on session memory, and to prioritize user confirmation for critical decisions like naming and role assignment.

onboarding