@maud-L could you please review the proposed FR translations for the 21 terms extracted from the GitLab Duo Agent Platform section (#937). For each term: confirm the FR is correct, or leave a comment with a correction. To reject a candidate entirely, note that too. Once you've reviewed, corrections will be transferred to !31. **TBX file**: `terminology_artifacts/duo_agent_platform-en.tbx` (MR !31) **Extraction issue**: #937 --- ## Terms needing confirmation The following terms have tentative FR proposals that need your review: | EN term | Proposed FR | Note | |---------|-------------|------| | flow | flow (borrowing) | Keep as English borrowing, or translate? "flux" exists as alternative. | | foundational agent | agent de base | Which adjective: "de base" or "fondamental"? | | foundational flow | flow de base | Follows rendering of "flow" and "foundational agent". | | managing project | projet gestionnaire | Noun+adjective pattern; "projet de gestion" is alternative. | | system prompt | prompt système | Borrowing "prompt système" or calque "invite système"? | | agent skill | compétence d'agent | Translate, or keep "skill" as borrowing? | | reachability analysis | analyse de joignabilité | No established FR term; "analyse d'accessibilité" is alternative. | | vetted MCP server | serveur MCP vérifié | "Vérifié" (verified) or "approuvé" (approved)? | ## Terms with higher confidence | EN term | Proposed FR | Note | |---------|-------------|------| | AI agent | agent d'IA | Quickterm Preferred | | custom agent | agent personnalisé | Follows rôle personnalisé pattern | | custom flow | flow personnalisé | Follows "flow" and "custom agent" | | external agent | agent externe | Direct translation | | AI Catalog | AI Catalog | DNT — product feature name | | version pinning | verrouillage de version | Standard software distribution equivalent | | GitLab Credits | crédits GitLab | Brand name retained; "crédits" lowercase | | composite identity | identité composite | Direct translation | | prompt injection | injection de prompt | Follows injection SQL pattern | | knowledge graph | graphe de connaissances | Standard NLP/IS translation | | embeddings | embeddings | Borrowing; "plongements" exists but not used in product docs | | RAG | RAG | Acronym kept | | flow trigger | déclencheur de flow | Follows "flow" rendering; "déclencheur" is standard GitLab CI/CD term | --- ## Review checklist - [x] AI agent - `agent d'IA` - [x] flow - `flow` ~~(**open**: keep as borrowing, or translate to "flux"? MS uses "flux")~~ - [x] foundational agent - ~~`agent de base` (**open**: "de base" or "fondamental"?)~~ is incorrect, please replace by `agent par défaut`. - [x] foundational flow - ~~`flow de base` (**open**: follows "flow" and "foundational agent" decisions)~~ is incorrect, please replace by `flow par défaut`. - [x] custom agent - `agent personnalisé` - [x] custom flow - `flow personnalisé` ~~(**open**: follows "flow" decision)~~ - [x] external agent - `agent externe` - [ ] managing project - `projet gestionnaire` (**open**: "projet gestionnaire" or "projet de gestion"?) --> I would need more context, do you have the EN source? - [x] system prompt - `prompt système` ~~(**open**: borrowing "prompt système" or calque "invite du système"? MS uses "invite du système")~~ - [x] AI Catalog - ~~`AI Catalog` (DNT)~~, please use `catalogue d'IA` instead. - [ ] version pinning - `verrouillage de version` (Skip for now and see how the LLM behaves) - [x] GitLab Credits - ~~`crédits GitLab`~~ please use `GitLab Credits` instead. - [ ] composite identity - `identité composite` (Skip) - [x] prompt injection - `injection de prompt` - [ ] agent skill - `compétence d'agent` (**open**: translate or keep "skill" as borrowing?) (Skip for now and see how the LLM behaves) - [x] knowledge graph - `graphe de connaissances` - [ ] embeddings - `embeddings` (borrowing) (Skip for now and see how the LLM behaves) - [x] RAG - `RAG` (acronym kept) - [ ] reachability analysis - `analyse de joignabilité` (**open**: "joignabilité" or "accessibilité"?) (Skip for now and see how the LLM behaves) - [ ] flow trigger - `déclencheur de flow` (**open**: follows "flow" decision) (Skip for now and see how the LLM behaves) - [ ] vetted MCP server - `serveur MCP vérifié` (**open**: "vérifié" or "approuvé"?) (Skip for now and see how the LLM behaves)