The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided marketing briefs and destination URLs to generate UTM parameters. While this presents a surface for indirect prompt injection, the risk is minimized by the skill's strict enforcement of naming conventions.
Ingestion points: Natural language briefs provided by the user and the contents of the local utm-registry.json file.
Boundary markers: The skill does not explicitly define delimiters for user input, relying instead on its internal validation logic.
Capability inventory: Access to Read, Write, and Bash tools for file management and link processing.
Sanitization: The skill enforces lowercase normalization and explicitly strips all characters except [a-z0-9-] for UTM parameter values, which serves as a strong defense against instruction or command injection within those fields.
[COMMAND_EXECUTION]: The skill frontmatter enables the Bash tool. The workflow involves reading and writing a local utm-registry.json file. The primary security controls are the strict character stripping and normalization rules described in the naming conventions, which prevent untrusted user input from being used to construct malicious shell commands or corrupt the registry file structure.

utm-link-generator