The Agent Skills Directory

[SAFE]: The skill follows security best practices for ServiceNow integrations. It uses parameterized record retrieval and structured MCP tools to gather field service data without hardcoding credentials or exposing sensitive endpoints to untrusted external parties.
[PROMPT_INJECTION]: The skill processes external data from ServiceNow work notes, which is a common surface for indirect prompt injection. The risk is considered minimal due to the skill's specific purpose of generating a formatted summary for human review.
Ingestion points: The skill retrieves untrusted content from sys_journal_field (work notes) and wm_task tables in SKILL.md (Steps 1 and 6).
Boundary markers: The summary uses delimiters such as === TASK SIDEBAR === and explicit section headers like LATEST NOTES: to separate different data types.
Capability inventory: The skill uses Bash (native tool) and ServiceNow MCP tools (SN-Read-Record, SN-Query-Table).
Sanitization: No explicit sanitization or filtering of the work notes is defined in the prompt template before the information is displayed to the technician.
[COMMAND_EXECUTION]: The skill lists Bash as a native tool. This is used in the provided examples to demonstrate standard REST API interactions via curl. The command usage is restricted to documented ServiceNow endpoints and does not involve piping untrusted remote content into the shell.

fsm-sidebar-summarization