qa-testing-strategy
Warn
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The
scripts/property_contract_runner.pyutility uses theimportlibmodule to dynamically load and execute Python code from an adapter file specified via the--contractcommand-line argument. As noted in the skill's own documentation, this execution is not sandboxed, which could allow a malicious adapter file to perform unauthorized actions if the agent is directed to load a compromised path. - [COMMAND_EXECUTION]: The
scripts/test_property_contract_runner.pyscript utilizessubprocess.run()to execute shell commands for verifying the runner's functionality. While the script employsshlex.split()andshlex.join()for safe command construction, the capability to spawn subprocesses increases the attack surface of the skill. - [REMOTE_CODE_EXECUTION]: Documentation in
references/observability-driven-testing.mdandreferences/chaos-resilience-testing.mdsuggests installing thetracetestCLI tool by piping a remote installation script from GitHub directly into the shell (curl ... | bash). Although the source (kubeshop) is a recognized provider, this remains an insecure installation pattern. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data such as
JUnit XMLtest reports andJSONtelemetry data. The instructions suggest using an AI agent to triage and analyze these results, which could lead to indirect prompt injection if the test output is manipulated by an external actor. - Ingestion points: Processes test results and telemetry data as described in
references/quality-metrics-dashboard.mdandreferences/observability-driven-testing.md. - Boundary markers: The provided templates do not define explicit delimiters or instructions to ignore embedded commands within the ingested test data.
- Capability inventory: The skill provides scripts that can execute Python code (
scripts/property_contract_runner.py) and spawn subprocesses (scripts/test_property_contract_runner.py). - Sanitization: No evidence of sanitization or filtering of external test result content before processing was found.
Audit Metadata