t-invest
Pass
Audited by Gen Agent Trust Hub on Jul 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted external data such as market news, analyst forecasts, and trading signals from the T-Invest API. This content could potentially contain adversarial instructions designed to influence the agent's trading behavior (Indirect Prompt Injection).
- Ingestion points: Market news, analyst forecasts, and trading signals fetched via the
news,forecast, andsignalscommands of the CLI. - Boundary markers: The skill does not explicitly define delimiters to separate external text from agent instructions, though it mandates the use of structured JSON for data parsing.
- Capability inventory: The agent has the ability to execute financial trades and set stop-orders using the
scripts/tinvest.cjstool. - Sanitization: No explicit sanitization or filtering of the text content from these external sources is mentioned.
- Mitigation: The skill significantly reduces risk by requiring explicit user confirmation for every transaction, showing a detailed preview of the order, and providing factual data checks (due diligence) before execution.
- [COMMAND_EXECUTION]: The skill operates by executing a bundled Node.js script (
scripts/tinvest.cjs) to interact with the T-Invest API. It also provides instructions to the user for environment setup, such as installing Node.js via system package managers (e.g.,apt,brew), which ensures that administrative actions are transparent and require user consent. - [DATA_EXFILTRATION]: The skill interacts with the official T-Invest API. It adheres to security best practices by directing the user to store sensitive API tokens in a local
.envfile (~/.config/tinvest/.env) rather than inputting them directly into the chat, which prevents secrets from being stored in the conversation history.
Audit Metadata