kotlin-gradle
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes project-specific data by reading Gradle build scripts (.gradle, .gradle.kts), configuration properties, and version catalogs (libs.versions.toml).
- Ingestion points: Project build files are scanned by scripts/inspect-project.py to extract plugin versions, targets, and compiler options.
- Boundary markers: The instructions do not define explicit prompt delimiters for the processed content.
- Capability inventory: The agent can execute the Gradle wrapper (./gradlew) and the bundled Python script.
- Sanitization: scripts/inspect-project.py includes an explicit redaction mechanism for property values whose keys contain sensitive fragments such as password, secret, token, credential, or user.
- [COMMAND_EXECUTION]: The skill performs project analysis and validation by executing shell commands.
- Evidence: Calls to python3 scripts/inspect-project.py and various Gradle wrapper tasks like ./gradlew projects, ./gradlew tasks, and ./gradlew check are used to verify build health.
- [DATA_EXPOSURE_EXFILTRATION]: The skill includes documentation regarding build reports and warns about the potential for secret exposure.
- Evidence: references/caches-daemon-reports.md and SKILL.md (Step 6) explicitly instruct the agent to avoid printing HTTP credentials and to redact environment variables from verbose reports.
Audit Metadata