codebase-to-course

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill mandates cloning and reading codebases and embedding "original code exactly as-is" into code↔English translation blocks, which forces any hardcoded secrets or API keys present in the repository to be output verbatim, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.80). The skill's SKILL.md explicitly instructs cloning arbitrary GitHub links ("If the user provides a GitHub link, clone the repo first (git clone <url> /tmp/<repo-name>)") and then to "read through the code" to build the course, meaning the agent fetches and interprets untrusted, user-generated public repo content that can materially influence its outputs and actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly instructs cloning a user-provided GitHub repo at runtime (e.g., git clone https://github.com/user/repo /tmp/) so the fetched repository contents are read into the agent's context and directly determine the generated course content, which meets the criteria for a runtime external dependency controlling the agent's outputs.