qa
Warn
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes various shell commands such as git, ls, grep, lsof, and curl. It also starts development servers using project-specific commands like npm run dev or npm start.
- [REMOTE_CODE_EXECUTION]: Executes code via project-defined test layers like npm test and supabase test db. It also generates and executes dynamic Node.js/Playwright scripts at the repository root.
- [EXTERNAL_DOWNLOADS]: Fetches data from external sources, specifically downloading test plans from URLs found in Linear attachments using curl.
- [DATA_EXFILTRATION]: Reads sensitive files like Playwright authentication states (storageState JSON files) to reuse sessions for automated browser testing.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from multiple sources to determine its actions.
- Ingestion points: Linear issue descriptions, acceptance criteria, attachment contents, and external test plan files.
- Boundary markers: The instructions do not define clear boundaries for external data; the agent is instructed to use this data as its primary checklist.
- Capability inventory: Significant capabilities are available, including file system access, network operations, and shell command execution.
- Sanitization: There is no explicit sanitization or validation of content extracted from external sources.
Audit Metadata