codebase-exploration

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow delegates repo recon to read-only subagents that grep/read files from the target codebase (e.g., files in an external repository), and those file contents are then summarized into a key-files map that the orchestrator uses—so outsider-authored free text from the repo can enter the LLM context via the subagents’ file reads.