The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes records from external ServiceNow CMDB data stores, introducing a vulnerability surface where untrusted data could attempt to influence the agent pipeline.
Ingestion points: Data is pulled into the agent context in SKILL.md via SN-Query-Table and SN-List-CmdbCis from fields within the cmdb_ci, cmdb_rel_ci, cmdb_ci_service, and service_ci_assoc tables.
Boundary markers: Absent. No protective isolation delimiters or system instructions exist to enforce that content retrieved from the configuration management database is handled purely as data rather than instructions.
Capability inventory: The skill utilizes powerful modification and administrative tools in SKILL.md, including SN-Execute-Background-Script (for server-side JavaScript evaluation), SN-Create-Record, and SN-Update-Record.
Sanitization: Absent. Content fetched from relationship tracks and CI descriptors is passed directly into text blocks and output routines without validation filters or escaping.

impact-analysis