literature-review-tools

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly tells the agent to ask the user for API keys and then run litrun.py env --set KEY=VALUE, which requires the LLM to receive or emit secret values verbatim (exfiltration risk).

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

• Third-party content exposure detected (medium risk: 0.30). In litrun.py the outsider-supplied --question (or --query) string is passed into the runtime workflow steps (e.g., paper-qa in workflows.json) without selecting any specific outsider-authored item first, so the agent ingests free text directly.

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

• Potentially malicious external URL detected (high risk: 0.90). The launcher clones and runs upstream web UIs at runtime (git clone + install + run), specifically cloning and executing code from https://github.com/assafelovic/gpt-researcher and https://github.com/stanford-oval/storm, which causes remote code to be fetched and executed locally.