Guardrails & Prompts - CruqAI

Guardrails

Guardrails are safety policies checked against agent input and output.

kind: pii, regex, keyword, max_length, or json_schema
target: input, output, or both
action: block, redact, or warn

cruqai guardrails apply block-pii --kind pii --target output --action redact
cruqai guardrails list

Prompts

A versioned registry of prompt templates. Templates use {{variable}} placeholders; every edit bumps the version so callers can pin a revision.

cruqai prompts apply support-triage \
  --template "You are a support agent. Answer: {{question}}"
cruqai prompts list

⌘I