zendesk-webhooks

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

• Third-party content exposure detected (high risk: 0.75). Yes—at runtime the webhook handler parses outsider-authored HTTP request body JSON (Zendesk-sent payload) into payload (JSON.parse(rawBody...) / json.loads(raw_body)), which is then available as text/fields in the handler logic and could be forwarded into an LLM context if later used.

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

• Secret detected (high risk: 1.00). I scanned the skill docs and examples for literal, high-entropy credential strings. I flagged the repeated base64 string "dGhpc19zZWNyZXRfaXNfZm9yX3Rlc3Rpbmdfb25seQ==" because it is a concrete signing secret (the static test signing secret Zendesk uses for test deliveries) and is present as a literal in multiple files and tests — it can be used to generate valid test signatures. I ignored obvious placeholders and examples such as "your_zendesk_signing_secret_here" and environment variable names, which are documentation placeholders and not real credentials.