Glossary
Each term has its French equivalent, so both language versions of the labs use the same words.
Lifecycle
| Term |
Definition |
In French |
| Brief |
The short product description (specs/idea.md) that every agent follows: idea, users, scope, out-of-scope, run rules. |
énoncé du produit |
| Prototype |
A clickable mock-up used to check the user experience before any real code exists. |
prototype |
| PRD |
Product requirements document: what the product must do and how success is measured. |
document d'exigences produit (PRD) |
| Technical specification |
The architecture, data model and contracts that turn the PRD into buildable work. |
spécification technique |
| ADR |
Architecture decision record: one short note per important technical decision and why it was made. |
fiche de décision d'architecture (ADR) |
| Slice |
A small, independently testable part of the app that the agents build in one task. |
tranche |
| Task |
One atomic unit of work with an owner, dependencies and required gates. |
tâche |
| Delivery |
Shipping the approved artifact to users; in the lifecycle it is the Deploy phase. |
livraison |
Quality and governance
| Term |
Definition |
In French |
| Quality gate |
An automated or reviewed check that must pass before a task counts as done (build, lint, unit, acceptance, review, security). |
passerelle de qualité |
| Human sign-off |
The mandatory approval by people (product owner, security team, tech lead) before anything is deployed. |
approbation humaine |
| Changes requested |
A sign-off outcome that sends the work back to Build and resets the approval. |
modifications demandées |
| Waiver |
A recorded human decision to accept a check that no agent can verify, with name, criteria and reason. |
dérogation |
| Critic review |
An independent review of the code for blocking defects. |
revue critique |
| Acceptance criteria |
The testable statements that say when a feature is done. |
critères d'acceptation |
| Smoke test |
A very short check that the deployed app starts and works. |
test de fumée |
| Rollback |
Returning to the previous working version. |
retour arrière |
| Retry |
A counted new attempt after a failed gate, made by the owning agent. |
reprise |
| Term |
Definition |
In French |
| Agent |
An AI specialist (designer, architect, developer, tester, and so on) with one role in the team. |
agent |
| Skill |
A packaged set of instructions an agent loads for one job, such as ait-init. |
compétence (skill) |
| Plugin |
The installable package that brings the agents and skills into Copilot. |
plugin (module d'extension) |
| Marketplace |
The registry Copilot reads to find and install plugins. |
marché (marketplace) |
| Orchestrator |
The agent that splits the work, dispatches the others and enforces the gates. |
orchestrateur |
| Scribe |
The agent that merges what the others wrote into the decisions and change log. |
scribe |
| Tracking store |
The .copilot-tracking/<run-id>/ folder that holds the state of a run. It is git-ignored. |
registre de suivi |
| Run |
One execution of the lifecycle, identified by a run id. |
exécution |
| Checkpoint |
A Git tag that marks the known-good state at the end of a lab. |
point de contrôle |
| Inbox |
The folder where each agent drops its own result file so parallel agents never overwrite each other. |
boîte de réception |
| Credit |
The unit of Copilot usage that a run consumes. |
crédit |