> ## Documentation Index
> Fetch the complete documentation index at: https://docs.forge.shieldtp.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Glossary
> Definitions for the main Forge products, controls, and workflow terms.
**Account context**
Customer-specific facts and guidance used to interpret work for that account.
**Action Plan**
Forge's structured understanding, investigation, and recommended steps for a ticket.
**Agent Library**
The area where users find and configure Custom Agents.
**Agent Runs**
The area where users inspect Custom Agent execution, output, failures, and follow-up work.
**AI BDR**
A human-operated queue of evidence-backed account-strategy proposals and their review state. Approval does not execute the external action.
**AI Engineer**
Forge's background helpdesk worker for eligible tickets.
**Ask mode**
The read-oriented interactive Chat mode. Protected reads may still require approval.
**Availability**
The organization setting that controls whether a tool can run in a product surface such as Chat, Custom Agents, Helpdesk, or Project Plans.
**Custom Agent**
A background workflow configured with instructions, tools, target, schedule, output, files, skills, memory, and optional judges.
**Dry run**
A Custom Agent run that skips write actions so its evidence and output can be reviewed safely.
**Full access**
The interactive Chat mode that can expose role-allowed read and write tools. It does not bypass other controls.
**Improvement**
A recurring helpdesk blocker or opportunity surfaced from ticket evidence. It does not automatically change a workflow.
**Judge**
An evaluation that scores or monitors a Custom Agent run against a defined criterion.
**PSA**
Professional services automation system. It remains the operational system of record for provider-backed ticket, project, time, billing, and administration data.
**Runbook**
An organization- or account-specific helpdesk procedure that can guide people and AI Engineer.
**Scope**
The organization, account, ticket, project, or other object that anchors a Chat or workflow.
**Skill**
Reusable agent guidance for a defined task, with optional references, files, tests, and versions.
**Structured output**
A defined set of fields a Custom Agent must return for consistent review or downstream use.
**Tool**
A specific capability Forge can use in a connected or platform system, such as searching records or updating a ticket.
**Tool approval**
An interactive, one-time allow or deny decision for a particular tool call.
**Work Plan**
The phases, tasks, estimates, dependencies, owners, and statuses used to deliver a project.