fin-kg-benchmark-report

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill instructs the agent to read all YAML artifacts and embed "all data" as JSON verbatim into a single HTML report, which requires including any secret values present in those artifacts directly in the generated output (exfiltration risk).

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md’s required workflow reads YAML artifacts from the evaluation output directory and the runtime generator (scripts/generate_report.py) loads their contents and injects them into the HTML template as const REPORT_DATA = ... (JSON), so any outsider-authored YAML text would become LLM-readable context via the report’s embedded data.