Glossary

The words Mainworthy uses, each in a sentence or two, with the page that says more. The words are the ones you see in the app and in command output.

Actor #

Whoever does something: a person, an agent, a pipeline, or a rule of the policy. Every result, edit and decision names its actor. See people, agents and pipelines.

Agent #

An AI model that works through the agent tools with a token of its own. What an agent's edits and answers are worth depends on the policy. See let an agent in.

Answer #

What you give a question: for a failure, one of defect, test, expected, flaky, infrastructure, already failing on base, or mistake. An answer that the policy accepts becomes a decision. See decide in the queue.

Attempt #

One try of a pipeline job on a commit. A retry is a higher attempt. See report results.

Basis #

Which version of which case a result is for: exact, assumed or unresolved. See how Mainworthy works.

Capability #

One thing a grant lets an actor do: read, record, produce, answer, plan or admin. See people, agents and pipelines.

Case #

What the software should do, written as a title and steps with expected results. A case has a ref such as C-12 and a history of versions. See write cases.

Change #

A commit or pull request that results were reported for, shown on the Changes screen with its gate. See changes and the gate.

Clause #

One row of a policy: for a kind of question under some conditions, who must answer and what an agent's answer is worth. See policy.

Decision #

A recorded answer that settles a question, with who gave it, on what evidence and under which clause. Only a decision changes what is current. A decision can be withdrawn, and both stay in the record.

Draft #

What you have typed into a case and not yet recorded as a version. A draft is kept on your device, per person. See write cases.

Finding #

Something a review noticed about a case, reported by a person or an agent. A finding is a question until someone accepts, rejects or defers it.

Gate #

The answer to "is this tested?" for a commit, a change or a run. It has four verdicts. mainworthy gate prints it and exits with its code. See changes and the gate.

Grant #

The list of capabilities an actor holds on one project, or on every project. See people, agents and pipelines.

Invitation #

A link an admin hands a person so that they can choose a password or register a passkey. It works once and runs out. For a person who already exists it is a recovery: accepting it replaces every way they signed in. See people, agents and pipelines.

Job #

A named part of a pipeline, such as e2e. A planned run can require that named jobs have reported.

Method #

How a case may be exercised: by a person, by an automated test, or by an agent. A case declares whether people and agents may exercise it, an automated test is admitted by its link to the case, and a run says which methods it asks for.

Passkey #

A key that a device or a password manager holds for one site and unlocks with a fingerprint, a face or a PIN. A person signs in with it in place of a password. It belongs to the host name of the instance.

Pipeline #

A CI system that reports results with a token of its own. A pipeline records results, says when a change ended, and reads the gate, and may be let read. It never edits, answers, plans or administers. See report results.

Planned run #

A run you start by hand: a set of cases, each pinned at its current version in a slot. See planned runs.

Policy #

How strict a project is: the table of clauses in force. loose, standard and strict are the three presets. See policy.

Project #

One suite with its own cases, results, queue and policy. An instance can hold several. A new instance starts with one called scratch.

Proposal #

An edit to a case, a link or the policy, as it was submitted. Saving and proposing are the same call. The policy decides whether a proposal takes effect at once or waits for a person. A proposal has a ref such as P-4.

Question #

Something that needs an answer: a proposed edit, a test with no case, a failed result, a finding, or closing a run. A question has a ref such as Q-7 and waits in the queue. See decide in the queue.

Queue #

The one list of open questions for a project, most blocking first. Open it with g q.

Result #

One observation of one test or case, such as a pass or a fail, with who produced it and its basis.

Rule #

The policy acting by itself. Where a clause says a rule answers, the answer is given at once and recorded as a decision by the rule.

Run #

A request for evidence: which results count and what is needed before it is complete. A change run exists for each reported commit. A planned run is started by hand.

Session #

One browser that is signed in. It ends at sign-out, after 14 days without use, after 90 days, and when what it was opened with (a token, a password or a passkey) stops working. A person lists and ends their own.

Slot #

One case in a planned run, pinned at a version, waiting for a result by a method the run asked for.

Tag #

A word on a case, written #checkout in the outline. The sidebar lists tags, and a run can pick cases by tag.

Test #

An automated test, as a runner reported it. A test has a ref such as T-9 and may be linked to a case.

Token #

A secret that names an actor and carries its grants. It is shown once, when it is made. See people, agents and pipelines.

Verdict #

What a gate says: passed, failed, blocked or pending. See how Mainworthy works.

Version #

One recorded state of a case's content. Every edit to the content that takes effect makes the next version. A move, a tag or archiving is recorded in the case's history and makes no version. You can compare two versions and restore an old one.