The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external graph data that could potentially contain malicious instructions.
Ingestion points: The skill reads graph nodes, descriptions, and connection metadata via iskron_orient and iskron_look (SKILL.md).
Boundary markers: There are no explicit instructions to the agent to treat data from the graph as untrusted or to ignore instructions embedded within the graph node bodies.
Capability inventory: The skill possesses the ability to modify the graph structure and metadata using iskron_arrow (SKILL.md).
Sanitization: No sanitization or validation of the graph content is described before the agent acts upon the data.

weaving