The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon 'Experience' content retrieved from external OpenViking URIs. If these stored memories contain adversarial instructions, the agent might follow them during high-stakes operations.
Ingestion points: The retrieval workflow in SKILL.md fetches content from the viking://~/memories/experiences root using environment-specific tools like find, search, and read.
Boundary markers: The skill includes explicit instructions to prioritize system/developer rules and user requests over Experience data. It also warns to ignore unsafe or conflicting guidance.
Capability inventory: The skill is used during 'executable, multi-step, or tool-based work' such as coding, data changes, configuration, and deployment, which involves significant system access.
Sanitization: There are no technical sanitization or validation steps mentioned for the retrieved content, beyond the agent's internal reasoning and the established priority order.

ov-experience-memory