gradle-build-performance

Fail

Audited by Gen Agent Trust Hub on Feb 15, 2026

Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION] (HIGH): The skill creates a high-risk surface for indirect prompt injection by having the agent process untrusted external data (build logs and scan results) while having the capability to execute commands. \n
  • Ingestion points: Build stdout/stderr and scan report data (SKILL.md).\n
  • Boundary markers: Absent; there are no instructions to help the agent distinguish between build output and commands.\n
  • Capability inventory: Shell execution via ./gradlew (SKILL.md).\n
  • Sanitization: None provided for the analyzed output.\n- [COMMAND_EXECUTION] (HIGH): The core workflow requires the agent to run ./gradlew commands. This capability is dangerous if the agent's logic is compromised by malicious logs or project configurations.\n- [DATA_EXFILTRATION] (LOW): The --scan diagnostic uploads build metadata, including environment details and dependencies, to scans.gradle.com. Users should be aware of this data exposure.\n- [CREDENTIALS_SAFE] (INFO): The configuration for remote build caches correctly uses environment variables (System.getenv) rather than hardcoded credentials.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 15, 2026, 10:31 PM