The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes user-controlled data from a persistent 'mind' graph, which constitutes an indirect prompt injection surface.\n- Ingestion points: Data is ingested through tools such as iskron_search, iskron_semantic_search, and iskron_look from the mind realm (SKILL.md §1).\n- Boundary markers: The instructions provide guidelines for the agent to distinguish between personal and project-related data and emphasize verifying the applicability of retrieved content before execution (SKILL.md, 'Приватность').\n- Capability inventory: The skill uses platform-native tools like iskron_update, iskron_add_karta, and iskron_add_phenomenon to manage the graph (SKILL.md §1b, §2). No shell execution or external network operations were detected.\n- Sanitization: The skill relies on agent evaluation and context-matching logic rather than automated data sanitization routines to mitigate risks from processed content.

minding