Content detection

Use this skill when

• You need to redact secrets embedded in arbitrary text (error messages, URLs, user input).
• You want to detect emails, JWTs, API keys, credit cards, or connection strings.
• You are writing custom detectors for application-specific patterns.
• You need to understand how overlapping detections are resolved.

Core model

Content detectors examine string values and return half-open ranges ({ start, end }) marking sensitive spans. Detected ranges are replaced with the configured replacement. Detectors do not run by default; they must be explicitly selected.

Detectors operate on:

• Primitive string values in the data graph
• Error name, message, stack, and cause fields
• The redact.text() method for standalone string scanning