Secra

• Scan Prompt — Scan text for prompt injection, jailbreak, persona hijacking, and data-exfiltration risk. Pass only the specific text you want checked — this tool never reads your conversation. Large inputs (e.g. a whole file) are handled automatically: they are split into parts, scanned, and merged into the single highest-risk verdict. Returns a verdict (allow/review/block), the detected threat type, a risk score, and a short recommendation.
• Sanitize Prompt — Return a cleaned version of the text with detected injection patterns neutralized. `level` is one of: light, moderate (default), aggressive. Pass only the specific text you want cleaned — this tool never reads your conversation. Large inputs are handled automatically: they are split, cleaned part by part, and stitched back together in order.

Schemas and invoke metadata: Directory API