codebase-to-demo

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill instructs the agent to clone and analyze codebases and to include "code exactly as-is" (and real integrations/configs) in the generated HTML, which would force the LLM to reproduce any embedded API keys, tokens, or plaintext secrets found in the repo or working directory — a direct secret-exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs the agent to fetch and clone external GitHub repositories and to "deeply analyze the codebase" (e.g., "git clone /tmp/" and "Read the README, the workflow files, the API calls" in SKILL.md/README), so it ingests untrusted, user-generated third‑party content that the agent must interpret and that materially guides its actions and outputs, creating a clear vector for indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs cloning user-supplied GitHub repositories at runtime (e.g., git clone https://github.com/user/repo) and then analyzes and may even launch that code (Playwright screenshots / runtime analysis), so fetched repo URLs are used at runtime and can both control model context/prompts and lead to remote code execution — flagging https://github.com/* (and equivalent git@github.com:... URLs).