Enhance auto-injected copilot instructions to have more relevant instructions for scope

Describe the feature or problem you'd like to solve

The applyTo glob pattern is not enough to scope down an instruction in large codebases with possibly hundreds of agent docs to read

Proposed solution

Support domain/category "tags" as an addition to  applyTo  glob patterns in  .instructions.md  frontmatter, so instructions can be auto-injected based on a semantic tag assigned to files, in combination with or independently of path-based glob matching. For example, a repo currently maintaining a separate, hand-curated library of convention docs (because many conventions are cross-cutting and don't map to a clean glob) could instead express every one of those conventions as a plain  .instructions.md  file tagged by concern (e.g.  error-handling ,  security ) — removing the need for a parallel documentation system just to work around glob's path-only scoping.

Example prompts or workflows

No response

Additional context

No response

Commencez par lire la gestion du frontmatter de .instructions.md et le comportement existant de correspondance des globs de applyTo. Suivez l’endroit où un scope est attribué aux fichiers et celui où les instructions sont injectées automatiquement ; done doit inclure des balises sémantiques fonctionnant indépendamment de la correspondance des chemins ou en complément de celle-ci, avec une couverture pour les exemples proposés de gestion des erreurs et de sécurité.