## What This Looks Like

A tool call appears to complete normally, but it also causes a downstream change the user does not see at the time. A record, metadata field, queue, workflow state, notification, integration, cache, or later processing step may change because of the tool call.

## Why Users Blame AI

The downstream effect is discovered after the AI action, so the AI looks responsible for a mysterious change. The deeper cause may be tool-side behavior, integration triggers, hidden metadata updates, automation watchers, shared state, or downstream systems that react to tool output.

## What to Check First

- Whether the tool has documented or hidden downstream effects
- Whether another workflow watches the changed object or field
- Whether metadata changes trigger automation
- Whether the tool call writes to shared state, not just the visible target
- Whether the final AI response mentions all tool effects
- Whether downstream consumers treat the tool result as authoritative

## When This Is AI-Adjacent

Use this AI-Adjacent Issue when a tool call causes hidden downstream change. If the downstream effect is a Workbench structural failure, use related Issues for undeclared side effects, downstream amplification, output handoff failure, or tool-result integration failure.