The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the use of f.run() for executing shell commands within workflows. It provides clear guidance and code examples emphasizing the importance of validating inputs and avoiding the interpolation of untrusted data into shell strings to prevent command injection.
[INDIRECT_PROMPT_INJECTION]: The documentation addresses the security implications of processing data from external triggers like webhooks. It explicitly recommends validating input schemas and field types before recording effects or using data in prompts, which mitigates the risk of indirect prompt injection.
[EXTERNAL_DOWNLOADS]: The skill references and depends on official vendor packages including @relayflows/surface and @relayflows/sdk, as well as official GitHub repositories from the author's organization. These are legitimate resources required for the functionality of the Relayflows engine.
[DYNAMIC_EXECUTION]: The provided test utility scripts/validate-input-example.test.mjs uses Node.js vm.runInNewContext to verify that code snippets extracted from the documentation are functional. This is a local developer tool for verifying documentation accuracy and operates on static local content.

writing-relayflows