GitHub Copilot adds semantic indexing to all workspaces and cross-repo search in @code

• pierceboggan's codebase-search thread says semantic indexing now works across all workspaces in @code, not just GitHub or Azure DevOps remotes, matching the VS Code 1.118 release notes.
• pierceboggan's cross-repo search post points to a new GitHub text search flow inside @code, and the official docs say it ships as a built-in githubTextSearch agent tool for grep-style search across a repo or an entire org.
• pierceboggan's Chronicle post introduces /chronicle and /chronicle:tips, while the release notes say Chronicle stores chat metadata, touched files, and external references in a local SQLite database.
• code's release post highlights a new Chat Customizations Evaluation extension, and the Marketplace listing says it analyzes .prompt.md, .agent.md, and .instructions.md files for contradictions, ambiguity, persona drift, and missing coverage.

You can read the full 1.118 notes, jump straight to the new cross-repo text search, and browse the prompt-eval extension. The weirdly useful bit is Chronicle: according to the official writeup, Copilot chat history is now queryable from a local SQLite index instead of just living as scrollback.

VS Code 1.118 expands the retrieval layer in two directions. The official release notes say semantic indexing for non-GitHub repositories is now rolled out to all users, so local and non-GitHub workspaces can participate in meaning-based search instead of literal string matching.

The same release adds a built-in githubTextSearch tool for exact-match lookups outside the current workspace, according to the GitHub text search section. That sits next to the existing semantic githubRepo tool, which means Copilot can now mix fuzzy retrieval inside a workspace with grep-style retrieval across repos or an entire org.

A separate 1.118 change, dedicated context for skills, is easy to miss but fits the same pattern. Skills can now run in an isolated subagent context so large tool outputs and reference material do not crowd the main chat.

The chat-history feature is more than a shortcut. The Chronicle notes say it records session metadata, conversation turns, files touched through tool calls, and references to PRs, issues, and commits in a local SQLite database.

That turns past Copilot sessions into something queryable. pierceboggan's post calls out /chronicle for history recall and /chronicle:tips for prompting suggestions based on prior sessions, while the release notes frame the feature as experimental behind github.copilot.chat.localIndex.enabled.

The prompt-eval piece shipped as a separate extension rather than a hidden release-note footnote. The Marketplace page says it is an LSP-based analyzer for .prompt.md, .agent.md, and .instructions.md files inside VS Code.

Its checks are unusually specific for first-party tooling:

• contradiction detection
• semantic ambiguity analysis with rewrite suggestions
• persona consistency checks
• cognitive load warnings for overly complex prompts
• semantic coverage checks for missing intent or error paths
• composition conflict analysis across prompt components

That makes this release about more than search. the @code announcement bundles retrieval, chat-memory recall, and prompt-quality linting into the same weekly update, which is a pretty direct sketch of where Copilot-in-VS-Code is heading.