java-injection-audit
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external evidence data labeled with EVID_* prefixes, typically generated by a separate pipeline (java-audit-pipeline). This architectural pattern creates an indirect prompt injection surface where maliciously crafted code comments or strings in the source being audited could be misinterpreted by the agent as authoritative evidence.
- Ingestion points: External evidence tags (EVID_*) referenced in SKILL.md and used in the Step 1 audit process.
- Boundary markers: None identified; the skill assumes the evidence tags provided by the environment are valid.
- Capability inventory: The skill provides logic for vulnerability assessment and scoring, which directs the agent's analytical focus.
- Sanitization: No instructions for the agent to sanitize or verify the authenticity of the incoming evidence data are provided.
Audit Metadata