spring-boot-in-action

Pass

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's components, including the instructional markdown and the pre-review Python script, are transparent and aligned with the stated educational and development goals. No obfuscation or remote access patterns were identified.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: Placeholder credentials (e.g., 'secret123' in examples/before.md and 'S3cr3tP@ss' in evals/evals.json) are used as examples of insecure practices that the skill is designed to flag and correct. They do not constitute a leak of real credentials.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze user-provided code, creating a surface for indirect prompt injection.
  • Ingestion points: User-supplied Java and property files ingested during the review process (SKILL.md).
  • Boundary markers: None. The skill does not define specific markers to delimit untrusted code from instructions.
  • Capability inventory: The skill can read local files via the provided scripts/review.py and agent file tools.
  • Sanitization: None. Inputs are processed directly via regex or NLP without sanitization.
Audit Metadata
Risk Level
SAFE
Analyzed
May 16, 2026, 01:22 PM
Security Audit — agent-trust-hub — spring-boot-in-action