The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data sources by querying ServiceNow tables such as wf_workflow_version and wf_context. This represents a standard vulnerability surface where maliciously crafted workflow metadata (e.g., names or descriptions) could be ingested into the agent's context.
Ingestion points: Data is retrieved from ServiceNow tables via snow_query_table and snow_workflow_manage.
Boundary markers: The skill instructions do not specify explicit delimiters to isolate external data from agent instructions.
Capability inventory: The skill allows creating activities (snow_create_workflow_activity), starting workflows (snow_start_workflow), and managing existing contexts (snow_workflow_manage).
Sanitization: No explicit sanitization or filtering of the content retrieved from the ServiceNow instance is documented in the instructions.
[SAFE]: The skill is authored by 'serac', which is consistent with the vendor infrastructure. All referenced tools and documentation are aligned with legitimate ServiceNow development and administration workflows. No evidence of credential theft, obfuscation, or remote code execution was found.

classic-workflow