ei-build-custom-deployment-blocks
Warn
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The provided Python (
build.py) and Node.js (build.js) skeletons use unsafe string interpolation to build shell commands. Values fromdeployment-metadata.json(such asinput_dir) are passed directly intoos.system()andexecSync()without sanitization (e.g.,os.system(f'cp -r {input_dir}/* {build_dir}')). This facilitates command injection if the input metadata is malicious. - [DYNAMIC_EXECUTION]: The templates demonstrate and encourage the use of shell command execution for the build process, which can be misused if the environment or inputs are compromised.
- [INDIRECT_PROMPT_INJECTION]: The skill's workflow involves processing external data that influences shell command execution, creating an injection surface.
- Ingestion points:
deployment-metadata.jsonis read by the build scripts. - Boundary markers: No delimiters or safety warnings are implemented for the ingested JSON data.
- Capability inventory: The scripts perform file system operations, command execution (
os.system,execSync), and archive creation. - Sanitization: The provided templates lack any validation or escaping of the metadata values before using them in sensitive operations.
Audit Metadata