AI-Adjacent Issues
AI-Adjacent Issues are problems users may experience as AI failures, but which may involve tools, apps, permissions, runtime behavior, configuration, product surfaces, or integrations.
These pages are separate from Workbench Issues. They point to related Workbench Issues when a structural diagnosis may apply.
Current AI-Adjacent Issues
-
Agent Keeps Adding Tool Dependencies
An agent keeps adding required tools, connectors, searches, files, APIs, or integrations instead of completing the task with the available surface.
ADJ-0006| Related Workbench Issues: 5 -
Agent Uses Old Tool Schema
An agent, prompt, workflow, or integration continues using an older tool schema after the tool contract has changed.
ADJ-0004| Related Workbench Issues: 5 -
AI Works in One Environment Not Another
The same AI task appears to work in one app, mode, model, account, workspace, or runtime but fails or behaves differently in another.
ADJ-0002| Related Workbench Issues: 5 -
Category Has Too Few Issues to Be Useful
A category, diagnostic area, or intake bucket has too few Issues, examples, or checks to support useful diagnosis or routing.
ADJ-0009| Related Workbench Issues: 4 -
Documented Field Not Available at Runtime
A documented field, option, parameter, property, or capability is expected by the user or agent but is not available in the actual runtime.
ADJ-0003| Related Workbench Issues: 4 -
Generated Output Changed Hidden Metadata
Generated output changes hidden metadata, identifiers, references, state, or system fields that affect later behavior but are not obvious in the visible output.
ADJ-0013| Related Workbench Issues: 5 -
Guardrail Applies to One Model Path Only
A guardrail, policy, validation rule, safety check, or workflow constraint applies to one model path, mode, runtime, or route but not another.
ADJ-0016| Related Workbench Issues: 5 -
Low-Level Configuration Overrides Workflow Policy
A low-level configuration, product setting, connector rule, runtime option, or environment flag overrides the workflow policy the user expects to govern the task.
ADJ-0015| Related Workbench Issues: 5 -
One Workspace Affects Another
State, memory, files, permissions, settings, tools, or workflow effects from one workspace affect another workspace where they should not apply.
ADJ-0008| Related Workbench Issues: 5 -
Permission Change Not Applied
A permission, approval, role, policy, or access change is made but the AI, agent, tool, or workflow continues behaving as if the old permission state still applies.
ADJ-0010| Related Workbench Issues: 5 -
Runtime Behavior Does Not Match Docs
Documented behavior, fields, tool schemas, permissions, or runtime capabilities do not match what actually happens during use.
ADJ-0005| Related Workbench Issues: 5 -
Too Many Workflows Depend on One Tool
Many workflows, agents, checks, or downstream paths depend on one tool, connector, API, or integration, creating a shared point of failure.
ADJ-0007| Related Workbench Issues: 5 -
Tool Call Caused Hidden Downstream Change
A tool call causes a downstream state, workflow, metadata, record, or integration change that is not visible at the point of action.
ADJ-0012| Related Workbench Issues: 5 -
Tool Effect Crosses Target Boundary
A tool, connector, function, or integration affects an object, file, record, workspace, or state surface beyond the intended target.
ADJ-0011| Related Workbench Issues: 5 -
Tool Failure Escalates Into Policy Decision
A tool, connector, API, or integration failure is treated as a policy, safety, approval, or governance decision instead of a runtime failure.
ADJ-0014| Related Workbench Issues: 5 -
Tool Not Supported in This Mode
A tool or capability appears unavailable in the current mode, environment, product surface, or runtime context.
ADJ-0001| Related Workbench Issues: 3