graphify

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 for Neo4j credentials if not provided and to replace placeholders (NEO4J_URI/NEO4J_USER/NEO4J_PASSWORD) with actual values in the generated code invocation, which requires embedding secret credentials verbatim in output.

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text can enter the LLM context via Step 3B “Semantic extraction (parallel subagents)”, where the Agent tool reads the user-provided corpus files (including docs/papers/images) and the subagent prompt instructs it to “Read the files listed” and extract entities/edges—if those files were authored by outsiders (e.g., cloned GitHub repos or downloaded webpages), their text is ingested into the subagent’s LLM context.