codebase-exploration

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

• Third-party content exposure detected (high risk: 0.70). This skill’s required workflow delegates repo “recon slices” to read-only subagents that grep/read source files from the target codebase, and those subagents’ returned file:line/path→role text is then merged into the main LLM context—so if the repo content is outsider-authored, it becomes outsider free text via the subagent→main-context path.