playwright
Warn
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONMETADATA_POISONING
Full Analysis
- [DYNAMIC_EXECUTION]: The run.js script creates a temporary file containing JavaScript code provided as a command-line argument, via stdin, or from a file path. It then executes this code using require(), which allows for the execution of arbitrary Node.js code without restriction.
- [INDIRECT_PROMPT_INJECTION]: The skill allows an agent to generate and run browser automation scripts based on user descriptions. This represents an attack surface for indirect prompt injection, as malicious data encountered during a browsing session could potentially influence the agent to generate and execute harmful scripts.
- [COMMAND_EXECUTION]: The run.js script and package.json configuration use execSync and npm scripts to perform installation tasks, such as npm install and npx playwright install. This involves executing shell commands to manage the environment.
- [DATA_EXFILTRATION]: The skill allows for the injection of custom HTTP headers into browser requests via environment variables (PW_HEADER_NAME, PW_HEADER_VALUE, PW_EXTRA_HEADERS). If sensitive credentials are stored in these variables, they could be exfiltrated to external domains during automation.
- [METADATA_POISONING]: The skill exhibits inconsistent authorship metadata, listing 'Claude Assistant' in SKILL.md and 'lackeyjb' in package.json, which differs from the provided context ('julianobarbosa').
Audit Metadata