add-product-e2e-eval
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the execution of shell commands using
pnpmandgitto manage the testing lifecycle, including repository location, type checking, unit testing, and running end-to-end evaluation suites. - [INDIRECT_PROMPT_INJECTION]: A vulnerability surface exists where the agent ingests data from external sources and local repository files while holding capabilities to modify source code and execute commands.
- Ingestion points: The agent reads existing repository documentation (e.g.,
README.md,FIXTURES.md,doc/evals.md) and interacts with external browser/API surfaces, including capturing screenshots and logs from production-like environments. - Boundary markers: There are no specific delimiters or instructions provided to isolate processed data from the agent's core instructions, potentially allowing embedded data to influence agent behavior.
- Capability inventory: The agent has the authority to write new test cases and registration code to files such as
catalog.tsandfixture-registry.ts, and can execute shell commands viapnpmas part of the test verification process. - Sanitization: Although the instructions advise sanitizing public fixtures and screenshots, there are no technical enforcement mechanisms or specific sanitization routines defined for the data processed during execution.
Audit Metadata