The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of the joelclaw command-line interface to submit workload artifacts and emit queue events.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data in the form of task descriptions and prompts which are subsequently processed by a background runner.
Ingestion points: The task.prompt field within the JSON request payload in SKILL.md serves as a primary input point for untrusted content.
Boundary markers: The skill uses a structured JSON schema to delimit task data.
Capability inventory: The skill provides functionality to invoke the joelclaw CLI via shell execution as seen in the provided Node.js example in SKILL.md.
Sanitization: The code examples utilize JSON.stringify to escape payload contents before they are passed as arguments to the spawn function.
[DYNAMIC_EXECUTION]: The skill provides a boilerplate implementation using node:child_process.spawn to dynamically construct and execute shell commands based on runtime input.

restate-workflows