The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided IP addresses and domain names to query infrastructure data, representing an attack surface for indirect injection via external data.
Ingestion points: User-supplied input strings (IPs or domains) provided during the conversation (SKILL.md).
Boundary markers: None explicitly defined in the prompt text; however, the workflow mandates classification and parameterization as a primary defense.
Capability inventory: subimageRunCypher for database query execution and subimageEnrichIp for external API enrichment (SKILL.md).
Sanitization: The skill explicitly enforces regex-based input classification and requires the use of parameterized Cypher queries ($value) to prevent injection (SKILL.md and references/cypher-templates.md).
[COMMAND_EXECUTION]: The skill uses tools to execute Cypher queries against an infrastructure graph database.
Evidence: The skill provides a library of templates in references/cypher-templates.md and strict instructions to use subimageRunCypher with mandatory parameterization and LIMIT bounds to prevent resource exhaustion or arbitrary query execution.

investigate-ip