senior-backend

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill includes examples that embed credentials directly in command-line arguments (e.g., an Authorization header with a bearer token and CLI --connection strings), which would require the LLM to include secret values verbatim in generated commands — an exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/api_load_tester.py, the required runtime workflow sends HTTP requests to attacker-controlled --urls and reads the full response body (response.read()), ingesting arbitrary outsider-authored free text from the target service at HTTPClient.request() time.