The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface where untrusted data provided during schedule registration is later processed by job handlers.
Ingestion points: The register method described in references/api.md accepts an input parameter of type unknown, which is stored and later passed to job execution logic.
Boundary markers: The library does not natively enforce boundary markers or delimiters between the schedule input and the agent's instructions.
Capability inventory: The job system (referenced in SKILL.md and references/api.md) includes the capability to execute process functions and submit further jobs via job.submit, potentially creating execution chains based on injected data.
Sanitization: The provided usage example in references/api.md demonstrates the use of the zod library for schema validation, which serves as a mitigation against malformed or malicious inputs if implemented by the developer.

sync-scheduler