openai-privacy-filter

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes examples that return raw detected spans (span.text) and an explicit API-key-like example ("sk-abc123xyz789"), which instructs the agent to read and output secret values verbatim (e.g., the audit_pii pattern logs detected PII without redaction).