taruvi-app-developer
Warn
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [CREDENTIALS_UNSAFE]: The deployment and export scripts (scripts/deploy-frontend.js and scripts/export-backend.js) are designed to receive the TARUVI_API_KEY as a command-line argument. This practice exposes the sensitive key to the system's process list, which can be viewed by other users or logged in shell history.\n- [DYNAMIC_EXECUTION]: The skill uses the manage_function tool to allow agents to author and update Python function bodies that are executed in the Taruvi runtime. This provides a mechanism for arbitrary code execution within the platform's sandbox environment.\n- [INDIRECT_PROMPT_INJECTION]: The serverless function entry point (def main(params, user_data, sdk_client)) and analytics queries process untrusted params dictionaries. While the documentation recommends using Jinja2 escaping and SQL placeholders, the ingestion of external data into execution contexts poses a vulnerability surface for indirect prompt injection.\n
- Ingestion points: Function parameters (params) in references/function-authoring.md and analytics query parameters in references/analytics-queries.md.\n
- Boundary markers: Explicit boundary markers are not documented for the params dictionary.\n
- Capability inventory: Functions have access to database operations, storage management, secret retrieval, and external network requests via the requests library.\n
- Sanitization: Documentation advises using %(name)s placeholders for SQL and {{ param }} Jinja2 escaping for analytics queries.\n- [COMMAND_EXECUTION]: The scripts/export-backend.js and scripts/check-versions.js files utilize Node.js child_process.execSync to perform system-level operations such as unzipping archives (unzip or PowerShell Expand-Archive) and checking package versions via npm view. While the arguments are currently based on hardcoded or server-returned values, the use of synchronous shell execution represents a potential risk if inputs were to be compromised.\n- [EXTERNAL_DOWNLOADS]: The skill includes scripts that fetch information from official registries and external APIs.\n
- scripts/check-versions.js fetches package metadata from the PyPI JSON API and the npm registry.\n
- references/function-scenarios.md includes an example of a function calling the OpenAI API at api.openai.com.
Audit Metadata