Glossary

Every term the app and this guide use, alphabetically. Each links to the page that explains it; Concepts puts them in the order the work happens.

Acceptance criteria — The claims a pull request must satisfy for its issue, listed on its verification page under "Does the PR do what the ticket says?", each verified with evidence or waived. See Pull requests.

Allow once — A one-time pass a person grants from a decision card when a loop needs to touch a protected path. See Needs-you inbox.

Auto-merge — Merging a pull request without a person once every gate is green and your policies allow it. See Pull requests.

Autonomy policies — The workspace rules that decide how much a loop may do on its own: auto-merge, human review, protected paths, spend guard and dry-run for new repos. See Policies.

Build Analyzer — The screen that mines your build history for what slows builds down and suggests changes. See Build Analyzer.

Build farm — The runners your workspace builds and tests on, on your own hardware. See Build farm.

Decision — A question a loop is blocked on, answered on its card in Needs You. See Needs-you inbox.

Dry run (workflow) — A rehearsal of a workflow against a sized issue in the Workflow Studio: no model is called, nothing runs and nothing is recorded. See Workflows.

Dry-run policy — The workspace setting under which pull requests open as drafts and nothing merges. See Concepts.

Evidence — What backs an acceptance criterion: a test result, a log, a measurement. See Pull requests.

Gate — A check a loop must pass to go on: a gate stage in a workflow, or a verification gate on a pull request. See Workflows and Pull requests.

Guardrails — The rules a run is held to — allowed paths, untouched CI config, a clean secrets scan — shown on its run console. See The run console.

Insights — Merge rate, cycle time, cost and where loops still need people, over a period you choose. See Insights.

Issue — A unit of work from your tracker, the input to a loop. See Issues.

Knowledge — What the loop knows about your repositories: skills, learned facts and playbooks. See Knowledge.

Learned fact — Something a loop learned about your code, kept once you confirm it. See Knowledge.

Live — A loop with dry-run off: a pull request that passes its gates and your policies merges without a person. See Concepts.

Loop — The cycle from issue to merged pull request, and the name a run carries in the app (Loop #N). See Concepts.

Mission Control — The dashboard: live loops, the queue, the build farm and what needs you. See Mission Control dashboard.

Model registry — The models your workspace allows, each under a name that routing refers to. See Models.

Model routing — Which model does which kind of work, with fallbacks. See Models.

Needs You — The inbox of decisions waiting for a person; its count is in the sidebar and the header. See Needs-you inbox.

Playbook — A reusable approach you can aim at any issue. See Knowledge.

Pool — A group of build-farm runners that work is sent to. See Build farm.

PR verification — The page that checks a pull request against its gates and acceptance criteria before it merges. See Pull requests.

Provider — A service that runs models — with the keys your workspace uses for it. See Providers.

Queue — The issues waiting for a loop, in order. See Issues.

Role — What a member may do in a workspace: owner, admin, member or viewer. See Roles.

Run — One pass of the loop over one issue, followed in the run console. See The run console.

Run console — The page for one run: stage timeline, agent transcript, changes, resources and guardrails. See The run console.

Runner — The agent you install on a build machine; it dials out to Ouroboros and runs builds and tests. See Build farm.

Simulated run — A run opened by the simulated-run driver for demonstration: no executor, repository or model did its work. See Concepts.

Sizing — The effort estimate, XS to XL, Ouroboros gives each issue. See Issues.

Skill — Instructions you write once for the loop to follow on every run. See Knowledge.

Stage — One step of a workflow — analyze, plan, implement, build, test, review — shown on a run's stage timeline. See Workflows.

Steer — A note you send a live loop from its run console to change its course. See The run console.

Test results — A run's test outcomes, flaky tests and triage. See Test results.

Ticket source — A tracker your workspace pulls issues from. See Sources.

Workflow — The stages a loop follows, edited on the Studio canvas or as code. See Workflows.

Workspace — Your team's tenancy: members, repositories, sources, policies and build farm. See Finding your way.