indexer

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt shows examples that embed sensitive values verbatim (e.g., viewingKey "mn_shield-esk1..." and sessionId strings) in GraphQL mutation bodies and subscription variables, which requires the LLM to output secrets directly and creates an exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime workflow fetches public on-chain data from the Midnight Indexer GraphQL endpoints (e.g., fetch(indexerUrl, { body: JSON.stringify({ query, variables }) })), and the LLM ingests the resulting GraphQL response fields (outsider-authored content from a public web service) as readable prose/JSON for further processing.