review-story

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly instructs the agent to read all files and to scan for hardcoded passwords/API keys and to produce "before/after code remediation" in its security report, which creates a high risk the LLM will include secret values verbatim in outputs (no redaction or secret-handling constraints are specified).

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

• Third-party content exposure detected (high risk: 0.85). Yes—Phase 2a “web-researcher” fetches public web content at runtime and the resulting research brief is passed into Phase 3/4 agent prompts as “Best Practices Context (from research)”, which is free-form outsider text.