The Agent Skills Directory

[CREDENTIALS_UNSAFE]: Hardcoded credentials are provided in multiple files for both Python scripts and CLI examples. Evidence in SKILL.md: The script servicenow_incident_manager.py contains AUTH = ("rest_admin", "SecureServiceNowPass123!") (line 50) and a cURL example uses -u "rest_admin:password" (line 125). Evidence in gemini.md: The CMDB health auditor script contains AUTH = ("rest_admin", "SecureServiceNowPass123!") (line 51) and a cURL example uses -u "rest_admin:password" (line 105). Evidence in gpt.md: A cURL example contains -u "admin:password" (line 111).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from ServiceNow APIs, which can be an entry point for malicious instructions. 1. Ingestion points: Data is retrieved from the ServiceNow Table API in SKILL.md (lines 53 and 67) and gemini.md (line 64). 2. Boundary markers: No delimiters or instructions to ignore embedded commands are used. 3. Capability inventory: The skill enables executing shell commands, managing Windows services, and making network requests. 4. Sanitization: No sanitization or validation of the retrieved API data is performed.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute system-level commands and manage services. Evidence in SKILL.md: Commands include start-service.bat and Test-WSMan (lines 118-121). Evidence in gemini.md: Includes Restart-Service (line 109). Evidence in gpt.md: Includes snc record create (line 107).

servicenow-desktop