webapp-testing

Warn

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
  • [COMMAND_EXECUTION]: The scripts/with_server.py script executes arbitrary shell commands provided via the --server CLI argument using subprocess.Popen(shell=True). This capability is intended for starting development servers but presents a significant risk if the agent is manipulated into executing malicious payloads.\n- [DYNAMIC_EXECUTION]: The skill uses subprocess.run and subprocess.Popen to dynamically execute server commands and testing scripts based on runtime parameters, which can lead to unauthorized code execution.\n- [METADATA_POISONING]: Instructions in SKILL.md (e.g., 'DO NOT read the source until you try running the script first') discourage the agent from auditing the underlying script logic. This concealment pattern reduces transparency and can be used to hide malicious intent.\n- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external web content, creating a vulnerability surface where malicious web pages could influence agent behavior.\n
  • Ingestion points: Web page DOM content and console logs are read into the agent's context in examples/element_discovery.py and examples/console_logging.py.\n
  • Boundary markers: None present; the agent is not instructed to treat web content as untrusted data.\n
  • Capability inventory: Includes shell command execution (scripts/with_server.py) and file system access for logs and screenshots.\n
  • Sanitization: No sanitization of ingested HTML or console messages is performed before they are used to determine subsequent actions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 16, 2026, 12:01 PM