cocoindex

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly tells the agent to ask the user for missing LLM API keys and to use real API key values in flows/examples (including patterns that place secrets into .env or pass them to functions like add_transient_auth_entry), which forces the LLM to handle/output secret values verbatim and creates exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.70). The required runtime workflow reads source document text via configured sources (e.g., flow_builder.add_source(cocoindex.sources.LocalFile(...)) / S3 / Postgres) and then passes that extracted prose into LLM-ingesting transforms like cocoindex.functions.ExtractByLlm(..., instruction=...), so if the source content is outsider-authored it becomes LLM context (indirect prompt injection risk).