The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the raisindb command-line interface for tasks such as raisindb deploy, raisindb package validate, and npx raisindb flow doctor. These are standard development and deployment tools for the RaisinDB platform.
[EXTERNAL_DOWNLOADS]: The skill references the raisindb tool via npx, which downloads and executes the package from the official npm registry. This is a standard practice for the vendor's own tooling.
[PROMPT_INJECTION]: The RaisinDB workflow engine supports variable interpolation (e.g., ${input.quantity} or {{ steps.pick_candidates.day }}) within workflow definitions. This creates an indirect prompt injection surface if the workflow processes untrusted data that is later passed to an ai_agent step or used in a command.
Ingestion points: Flow input (input.*), step outputs (steps.<id>.*), and trigger data in SKILL.md.
Boundary markers: None explicitly provided in the workflow templates to delimit untrusted data from instructions.
Capability inventory: Subprocess execution via raisindb deploy and network operations via API endpoints (/api/flows/, /api/inbox/).
Sanitization: The documentation suggests using flow doctor for structural and logic validation, though it does not detail content-level sanitization for interpolated data.

raisindb-workflows