The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a bundled Node.js CLI tool (cli.mjs) to perform API requests against ServiceNow. It supports various subcommands for reading tables, fetching attachments, and inspecting schemas.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from external ServiceNow instances (e.g., ticket comments, attachment metadata, and service catalog variables).
Ingestion points: Data is pulled via the Table API, Attachment API, and Service Catalog API endpoints.
Boundary markers: No specific delimiters for ServiceNow content are defined in the provided instruction files.
Capability inventory: The skill can execute shell commands via node cli.mjs and write files to the local system using the --out flag in the attach command.
Sanitization: The SAFETY.md documentation includes guardrails requiring the agent to inspect batch JSON files and refrain from piping output to network commands, providing a layer of human-in-the-loop and behavioral sanitization.
[DATA_EXFILTRATION]: The instructions and the SAFETY.md file include specific prohibitions against piping CLI output to network-transmitting commands (e.g., curl, nc, ssh), significantly reducing the risk of exfiltrating sensitive organizational data.
[CREDENTIALS_UNSAFE]: The skill follows security best practices by directing users to store ServiceNow credentials (domain, username, password) in a local .env file rather than hardcoding them or passing them via insecure command-line history.

servicenow-agent-hardened