codebase-analyzer

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt requires the agent to read and cite specific configuration/manifest/CI files (e.g., envs, Dockerfiles, workflow files) and to reference file contents as evidence, which can force the model to reproduce secret values verbatim from the repository in its generated analysis.

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime workflow ingests a user-supplied directory path and then reads files from that directory (e.g., README.md, source files, docs) into the LLM context for analysis; if that directory contains outsider-authored free text (e.g., third-party docs/issues/wiki content), that text can be fed into the agent’s LLM context via the file-reading step.