The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill utilizes the SN-Execute-Background-Script tool to run arbitrary JavaScript on ServiceNow instances for validation (Step 1.2), checking retrieval status (Step 2.3), bulk resolving conflicts (Step 4.3), and rollback planning (Step 7.2). While these specific implementations are benign and follow administrative best practices, the tool itself provides a high-privilege remote execution surface.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from ServiceNow environment records.
Ingestion points: Data is ingested through SN-Query-Table and SN-Inspect-Update-Set in Step 1.1, Step 3.2, and Phase 6, where the agent reads update set payloads, logs, and preview problems.
Boundary markers: The instructions lack delimiters or explicit 'ignore instructions' warnings when handling content from retrieved ServiceNow records.
Capability inventory: The agent has extensive capabilities, including server-side script execution (SN-Execute-Background-Script), record creation (SN-Create-Record), and record updates (SN-Update-Record).
Sanitization: There is no evidence of sanitization or filtering applied to the data retrieved from the sys_update_xml or syslog tables before the agent acts upon it.

deployment-workflow