diff --git a/src/wizard/onboarding.ts b/src/wizard/onboarding.ts index 764274ef9..782e28908 100644 --- a/src/wizard/onboarding.ts +++ b/src/wizard/onboarding.ts @@ -54,8 +54,8 @@ async function requireRiskAcknowledgement(params: { const danger = (value: string) => colorize(rich, theme.error.bold, value); const warning = (value: string) => colorize(rich, theme.warn.bold, value); const good = (value: string) => colorize(rich, theme.success.bold, value); - const subtle = (value: string) => colorize(rich, theme.warn, value); - const alert = (value: string) => colorize(rich, theme.error, value); + const emphasis = (value: string) => colorize(rich, theme.accentBright.bold, value); + const alert = (value: string) => colorize(rich, theme.error.bold, value); await params.prompter.note( [ @@ -63,22 +63,24 @@ async function requireRiskAcknowledgement(params: { "", alert("Prompt injection can override your bot. This risk is real and not solved."), "", - subtle( + emphasis( "Untrusted content can carry hostile instructions (links, web pages, files, logs, pasted text).", ), - subtle("If tools are enabled, a bad prompt can leak secrets, run commands, or access files."), + emphasis( + "If tools are enabled, a bad prompt can leak secrets, run commands, or access files.", + ), "", warning("Model policy (required for tool-enabled or untrusted input):"), `- ${good("Recommended:")} Anthropic Opus 4.5.`, `- ${warning("Not recommended:")} weaker tiers like Sonnet or Haiku.`, `- ${warning("Supported but not recommended:")} free models, cheap models, or local models.`, "", - warning("Required safeguards:"), + emphasis("Required safeguards:"), "- Pairing/allowlists + mention-gating.", "- Sandboxing + minimal tool access.", "- Keep secrets out of the agent's reachable filesystem.", "", - warning( + emphasis( "Run this regularly (especially after changing config or exposing network surfaces):", ), "clawdbot security audit",