code-graph

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly supports cloning and indexing remote public repositories (see SKILL.md example "cgraph index-repo https://github.com/org/repo") and tells the agent to read source files ("use the Read tool to load the actual source code"), so it ingests untrusted, user-generated web content that can influence queries and actions.