Glossary

The words this workspace uses, in alphabetical order. Each entry links to the page that explains it in full.

#Acceptance criteria

Two lists on every ticket that say what "done" means: plain criteria the client signs off, and technical criteria the reviewer checks. Agents are given both, and both are shown again at the gate. See Tickets and acceptance criteria.

#Agent

A teammate on the bench: a name, a role, its own instructions and an AI model. Agents write code with Claude Code or Codex. See AI coding agents.

#Allow shell commands

An option on the assign dialog that lets an agent run commands, such as installing packages or running tests, for that one run. Off by default.

#Approval gate

The point where every piece of agent work stops and waits for a person. Nothing goes further until someone approves it. See Runs and the approval gate.

#Approver

Someone whose group allows them to approve agent work, or an admin of that project.

#Assign to agent

The button on a ticket that starts agent work. It opens a dialog where you choose the agent, the mode and the options.

#Attachment

A file on a ticket or one of its comments. Agents can read PDFs and images; other files are kept for people.

#Awaiting approval

A run that has finished and is waiting at the gate. The ticket stays locked until someone decides.

#Awaiting permission

A run that stopped because it needs something it was not given. Nothing is lost while it waits, and granting the request lets it carry on.

#Baseline Standards

The shared rules every agent on the bench follows.

#Bench

The agents available to assign. Only active agents appear on the assign dialog.

#Branch

Each ticket has a branch of its own, named after the ticket. Agents work there, and every attempt at the ticket lands in the same place.

#Client

The organisation the work is for. Projects belong to a client. An enquiry from the contact page can be turned into a client.

#Client update

A short note for the client describing a change that has shipped. Drafted from the run and editable before it is sent.

#Connector

A tool Jarvis is connected to, such as GitHub, Jira or an AI model. Set up by an administrator. See Connectors.

#Constraint

A note category: something that must stay true. See Goals and notes.

#Context

A note category: background facts a newcomer would need.

#Decision

A note category: a choice that was made, and why.

#Define, Plan, Steer, Fund, Ship

The five jobs every delivery needs. Define sets what is being built; Plan breaks it into pieces; Steer keeps the work on course; Fund checks the effort is worth it; Ship gets it to the client. See How we deliver.

#Deploy

Sending approved work to your hosting, for projects set up to do so. Never happens without an approval.

#Enquiry

A message sent through the public contact page. Enquiries appear on the Enquiries screen for people allowed to see them, and can be turned into a client.

#Execution mode

How far a run may go: Analyse only, Prepare implementation plan, Implement in branch, or Implement, test and capture preview.

#External key

A ticket's number in the tool it came from, such as a Jira key, a Trello card or a GitHub issue. Shown in its own column and linked back to the source.

#Goal

The end state a project's current phase works toward, written so someone could check it is true. Shown on the project page and given to every agent. See Goals and notes.

#Group

What decides which screens and actions you can use. Your administrator puts you in one.

#Guidance

A note category: how the team prefers things done.

#Instruction

The optional note you write when assigning an agent, for anything the ticket does not say.

#Jarvis

The name of the workspace, and of its voice screen, which answers questions about the work out loud and takes simple actions. See Jarvis voice.

#Knowledge

Short documents attached to an agent, for every project or for one. The ones relevant to a ticket are given to the agent.

#Linked repositories

Other projects whose code this project's agent work also includes, on the same branch. Links go one way. See What an agent reads.

#Notes

Project writing in four categories (context, constraint, guidance, decision), kept on the Specs & notes tab. Agents read them all.

#Outcome

A lasting capability the system must have to reach the goal. Not a task and not a feature list. See Outcomes and quality checks.

#Outcome statuses

Proposed, Ready, In progress, Delivered and Retired. Ready, In progress and Delivered need all six quality checks ticked.

#Plan

A run that writes a plan instead of changing files, either because its mode asked for one or because the agent could not write code.

#Proposal Studio

The module for writing proposals for new work from a job post, using your portfolio and profile.

#Pull request

Where approved agent work goes for your team's usual code review. Its description is written by the agent as the change's author.

#Quality checks

The six tests an outcome must pass before it is Ready: an outcome not a task, distinct, testable, lasting, clear, and serves the goal.

#Request changes

A decision at the gate that records what is wrong, sets the attempt aside and moves the ticket to Changes requested.

#Retry

Starting a fresh attempt after a run failed, was cancelled or was rejected.

#Review comments

Comments on a pull request. Address review comments sends the new ones to the agent as its next instruction.

#Review with AI

A button on the Outcomes tab that asks an AI model to assess every outcome against the six checks and point out gaps and overlaps. Advice only.

#Run

One attempt by one agent at one ticket.

#Skill

A written procedure an agent can follow, kept in the skills library. Only published skills are given to agents.

#Specs

Project writing about the system itself, in categories such as stack, architecture, api, database, infra, integration and security. Kept on the Specs & notes tab.

#Sprint

A time box from your Jira board. Tickets can be moved between sprints from the ticket or by voice.

#Story points

A ticket's size. Half points are allowed.

#Stuck run

A run that has gone quiet. It can be continued, sent to the gate with what it already changed, or closed.

#Task log

A record of your day: hours and what you did, per project. Feeds the daily, weekly and consolidated timesheet reports. Visible to you and administrators only.

#Ticket

The unit of work. Also called a task.

#Trusted browser

A browser you chose to let skip the two-factor code for a while. You can remove it from your security page.

#Usage report

A summary of what agent work cost, by project, agent, model and person.

#Workflow

The rules for which status may follow which on the board.

#Worklog

One entry per day per ticket, with minutes, drafted from what happened that day.