Wind Alice Fact Check

The problem it solves

In research, trading, and everyday information review, you often need to judge whether a company’s revenue actually crossed a threshold, whether an industry ranking is accurate, or whether an event description matches public records. These claims usually arrive as mixed prose from news posts, research summaries, meeting notes, or other AI outputs. Manually checking every number is slow, while accepting the text at face value can hide timing, definition, and source issues. Wind Alice Fact Check turns this into a repeatable verification workflow.

How it works

The skill calls the Wind Alice Agent through a CLI. The host agent passes the original text or question as --prompt, while the CLI waits for the task to finish and returns agentResult.value on stdout. The core flow includes:
- Identify verifiable data points: extract revenue, profit, sales volume, market cap, price change, rankings, market share, and corporate events.
- Check each point: compare it against Wind financial data and authoritative public sources, then label it as accurate, inconsistent, or not found.
- Produce a structured report: list the original claim, verification result, and a brief explanation; flag possible definitional or timing issues and suggest next steps when data cannot be found.

Delivery should use the stdout body first, rather than expanding silently downloaded report attachments. If the agent asks for missing original text or other required input, the host agent should pass that question to the user and pause until the prompt is completed.

Boundaries and notes

This skill fits passages containing concrete numbers, rankings, or events; it is not intended for single-point questions like asking how much a stock changed today. Accuracy depends on how complete and specific the source text is. Before use, configure WIND_API_KEY and follow AGENT.md for execution constraints, avoiding duplicate calls, cross-session context pollution, or misreading Windows console encoding issues.