The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for using the raisindb CLI to validate and deploy workflows. Commands like flow doctor, package validate, and deploy are standard administrative tools for the RaisinDB platform.
[EXTERNAL_DOWNLOADS]: The use of npx raisindb initiates a download and execution of the RaisinDB CLI package from the public npm registry, which is the expected method for installing the vendor's development tools.
[INDIRECT_PROMPT_INJECTION]: The ai_agent step type allows flow data and step outputs to be processed by an LLM, creating a surface for indirect prompt injection.
Ingestion points: Untrusted data enters the flow via input.* and trigger.* namespaces and is passed to agents through properties like prompt and include_context.
Boundary markers: The workflow uses ${expr} and {{ expr }} for data interpolation, but the skill does not explicitly describe the use of safety delimiters or system instructions to mitigate prompt injection.
Capability inventory: The workflow engine is capable of executing server-side functions, creating human tasks in an inbox, and managing database state.
Sanitization: The documentation does not specify sanitization or validation routines for data being interpolated into agent prompts.

raisindb-workflows