The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill follows standard architectural patterns for ServiceNow integrations.
[PROMPT_INJECTION]: The skill defines instructions for building AI components that process potentially untrusted record data, which creates an indirect prompt injection surface. The documentation proactively addresses this by recommending validation steps in the troubleshooting section.
Ingestion points: Untrusted data enters the skill context through the sn_now_assist_skill_input table, specifically via the record_field and user_input source types.
Boundary markers: Prompts utilize curly brace delimiters (e.g., {incident_description}) to interpolate data into system templates.
Capability inventory: The skill provides functionality to modify records via SN-Create-Record and SN-Update-Record and lists access to a Bash shell tool.
Sanitization: The author explicitly suggests implementing input validation within a preprocessing prompt step as a mitigation strategy for potential injection vulnerabilities.

skill-kit-custom