The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from system tables that could be influenced by external factors, creating a vulnerability surface where malicious instructions could be embedded in data processed by the agent.
Ingestion points: The skill reads from sys_slow_transaction, syslog_transaction, and syslog tables (SKILL.md), which contain URLs and log messages that can be generated by external web requests.
Boundary markers: Explicit delimiters or 'ignore embedded instructions' warnings are absent when processing tool outputs.
Capability inventory: The skill utilizes SN-Execute-Background-Script (server-side JavaScript execution), SN-Create-Record (database write access), and SN-Query-Table (database read access) tools (SKILL.md).
Sanitization: There is no evidence of specific sanitization or filtering logic applied to the external log content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill instructs the agent to use the SN-Execute-Background-Script tool to run administrative JavaScript code on the ServiceNow server.
Evidence: Multiple phases (e.g., Phase 1.2, 3.4, 4.1, 9.1) provide scripts for server-side execution to analyze queries and system health.
[DYNAMIC_EXECUTION]: The skill generates and executes JavaScript code at runtime using platform-specific background script tools.
Evidence: Phases 1.2, 2.4, 3.4, 4.1, and 9.1 involve constructing JavaScript strings that are then passed to the SN-Execute-Background-Script tool for evaluation.

performance-optimization