configuring-exports

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime ingestion can include outsider-authored free text via WebhookExport listeners, where incoming HTTP request payloads (parsed via requestMediaType and emitted using pathToRecords) are passed into subsequent flow steps.

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I reviewed the flagged candidates against the files.
• The Stripe key (sk_test_51LZIr9B9Y6YIwSKx8647589JKhdjs889KJsk389) is present verbatim in an examples array for the webhook HMAC key field (references/schemas/webhook.yml). It is a high-entropy API key string (even though it is a test-prefixed Stripe key), and literal API keys should be treated as credentials and flagged.
• The JWT-looking string (eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9....SflKxwRJSMeKKF2QT4fwpMeJf36POk6yJV_adQssw5c) appears only as an illustrative example in references/schemas/distributed.yml. That particular token is the well-known JWT demo (payload contains "John Doe" and sample iat) used in documentation; it is a static demonstration value and not a live credential. Per the scanning rules, documentation-only demo tokens that are non-functional examples should be ignored as false positives.