atr-behavior
Warn
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill documents that
.test.propertiesfiles support shell command expansion using the$(command)syntax. This allows for arbitrary command execution on the host system whenever theatrtool reads these configuration files, which is a significant risk if the files are part of an untrusted codebase. - [DYNAMIC_EXECUTION]: The tool compiles natural-language
.test.txtspecs into.test.jsfiles at runtime. This dynamic generation of executable code from text files poses a security risk if the input specifications are maliciously crafted to include harmful logic. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted natural-language specifications (
.test.txt) which are processed by an LLM for compilation, repair, and failure triage. This creates an attack surface where malicious instructions embedded in the specs could influence the agent's actions or the generated code. - Ingestion points: Specification files (
.test.txt) and property files (.test.properties) processed by the agent and the ATR tool. - Boundary markers: No specific delimiters or safety instructions are mentioned to isolate the natural-language test content from the model's system instructions.
- Capability inventory: The environment allows shell command execution (via property expansion), browser automation, and network access for telemetry.
- Sanitization: The skill does not describe any sanitization or validation steps for the natural-language input before it is passed to the model for compilation.
- [DATA_EXFILTRATION]: The skill supports the use of an OTLP telemetry endpoint (
--otel-endpoint) to export traces, metrics, and logs. While intended for monitoring, this feature could be used to exfiltrate sensitive information, such as application data or environment variables captured in logs, to a remote server controlled by an attacker.
Audit Metadata