Update Learning Hub with VS Code 1.129 and CLI v1.0.71 features

- agents-and-subagents: add VS Code Agent Host (1.129) section with session-management tools; note /subagents model picker now preserves reasoning effort and context tier (v1.0.71); update platform table - building-custom-agents: add BYOK models usable with Copilot harness (VS Code 1.129); update model table - copilot-configuration-basics: add chat.agentHost.enabled setting and '!' terminal command prefix from VS Code 1.129 - creating-effective-skills: document prompt-file-to-skill migration (experimental, chat.customizations.promptMigration.enabled, VS Code 1.129) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-08-01 23:12:29 +00:00 · 2026-07-19 21:00:58 +00:00
parent 26fe2d126b
commit e4b43c9887
4 changed files with 65 additions and 7 deletions
@@ -3,7 +3,7 @@ title: 'Agents and Subagents'
 description: 'Learn how delegated subagents differ from primary agents, when to use them, and how to launch them in VS Code and Copilot CLI.'
 authors:
  - GitHub Copilot Learning Hub Team
-lastUpdated: 2026-07-01
+lastUpdated: 2026-07-19
 estimatedReadingTime: '9 minutes'
 tags:
  - agents
@@ -109,6 +109,28 @@ Then summarize the findings into one recommendation.

 By default, subagents do not keep spawning additional subagents. In VS Code, recursive delegation is controlled by the `chat.subagents.allowInvocationsFromSubagents` setting, which is off by default.

+### VS Code Agent Host (VS Code 1.129+)
+
+VS Code 1.129 introduced the **agent host** — a dedicated, long-lived process that runs Copilot, Claude, and Codex agent sessions. Enabling the agent host aligns VS Code's agent behavior more closely with Copilot CLI and the Copilot SDK, and unlocks **session-management tools** that let agents inspect and message sibling sessions.
+
+Enable it in settings:
+
+```json
+{
+  "chat.agentHost.enabled": true
+}
+```
+
+Once enabled, agents can use session-management tools to:
+- **List** other active agent-host sessions
+- **Read** the conversation history of a specific session
+- **Create** new sessions programmatically
+- **Send messages** to another session and receive its reply
+
+This unlocks a new class of orchestration patterns directly inside VS Code: a coordinator agent can spin up peer sessions for specialized work and collect their results — similar to how `/fleet` works in Copilot CLI, but within the VS Code environment. GitHub Enterprise repositories are also supported with the agent host.
+
+> **Note**: The agent host is a new feature in VS Code 1.129 and behavior may evolve. Check the [VS Code release notes](https://code.visualstudio.com/updates) for the latest details.
+
 ## Launch subagents in Copilot CLI

 In GitHub Copilot CLI, the clearest end-user entry point is **`/fleet`**. Fleet acts as an orchestrator that decomposes a larger objective, launches multiple background subagents, respects dependencies, and then synthesizes the final result.
@@ -144,6 +166,8 @@ In v1.0.64+, you can configure the rubber-duck agent (including its complementar
 /subagents          # open the subagents configuration panel
 ```

+In v1.0.71+, the `/subagents` model picker **preserves each agent's reasoning effort and context tier** when the panel is reopened. Previously, any per-agent reasoning or context-tier adjustments you made were reset on the next open; now your choices persist for the duration of the session.
+
 Or you can still enable experimental features and select it from the agent picker:

 ```
@@ -188,7 +212,7 @@ VS Code documentation describes both subagents and the `handoffs` frontmatter pr

 That means you should think about delegation features in product-specific terms:

- **VS Code**: supports subagent concepts, allowlists, and handoff-oriented agent composition
+- **VS Code**: supports subagent concepts, allowlists, handoff-oriented agent composition, and (1.129+) the agent host with session-management tools
 - **Copilot CLI**: exposes practical orchestration through commands like `/fleet`
 - **GitHub.com coding agent / cloud agent**: supports custom agents, but some VS Code-specific frontmatter is intentionally ignored

@@ -3,7 +3,7 @@ title: 'Building Custom Agents'
 description: 'Learn how to create specialized GitHub Copilot agents with custom personas, tool integrations, and domain expertise.'
 authors:
  - GitHub Copilot Learning Hub Team
-lastUpdated: 2026-07-09
+lastUpdated: 2026-07-19
 estimatedReadingTime: '10 minutes'
 tags:
  - agents
@@ -262,6 +262,10 @@ The agent can then query your database, analyze query plans, and suggest optimiz
 | Quick analysis, simple tasks | Claude Haiku or GPT-4.1-mini |
 | Large codebase understanding | Models with larger context windows |

+#### BYOK models with the Copilot harness (VS Code 1.129+)
+
+In VS Code 1.129, **Bring Your Own Key (BYOK) models** are now usable with the **Copilot harness** in the Agents window. This means you can point a custom agent at a BYOK model endpoint (e.g., your own Azure OpenAI deployment) and it will still benefit from Copilot's tool calling, skills discovery, and agent host infrastructure — rather than running as a bare chat completion. To use a BYOK model in an agent, configure the model endpoint under VS Code's language model settings and then reference the model identifier in your agent's `model` frontmatter field.
+
 ### Organizing Agents in Your Repository

 ```
@@ -147,6 +147,29 @@ Prevent Copilot from accessing specific files or directories.

 **Why it matters**: Exclude sensitive files, generated code, or dependencies from Copilot's context to improve suggestion relevance and protect confidential information.

+### Agent Host (VS Code 1.129+)
+
+The **agent host** runs Copilot, Claude, and Codex agent sessions in a dedicated long-lived process, aligning VS Code's agentic behavior with Copilot CLI and the Copilot SDK. It unlocks session-management tools that let agents list, read, and message sibling sessions, as well as GitHub Enterprise support for Copilot agents.
+
+```json
+{
+  "chat.agentHost.enabled": true
+}
+```
+
+**Why it matters**: Enables a richer orchestration model inside VS Code where agents can coordinate across sessions. See [Agents and Subagents](../agents-and-subagents/) for details on the session-management tools this unlocks.
+
+### Running Terminal Commands from Chat (VS Code 1.129+)
+
+Use the `!` prefix in the VS Code chat input to run a terminal command directly without leaving the chat:
+
+```
+! npm test
+! git status
+```
+
+**Why it matters**: Quickly inspect output (test results, lint errors, git state) mid-conversation without switching to a terminal panel.
+
 ## Repository-Level Configuration

 The `.github/` directory in your repository enables team-wide customizations that are version-controlled and shared across all contributors.
@@ -3,10 +3,7 @@ title: 'Creating Effective Skills'
 description: 'Master the art of writing reusable, shareable skill folders that deliver consistent results across your team.'
 authors:
  - GitHub Copilot Learning Hub Team
-lastUpdated: 2026-07-13
-estimatedReadingTime: '9 minutes'
-tags:
-  - skills
+lastUpdated: 2026-07-19
  - customization
  - fundamentals
 relatedArticles:
@@ -387,6 +384,16 @@ You can also run `/skill` (or the existing `/skills`) inside an interactive sess

 A: Skills replace the older prompt file (`*.prompt.md`) format. Skills offer agent discovery (prompts were manual-only), bundled assets (prompts were single files), and cross-platform portability via the Agent Skills specification. If you have existing prompts, consider migrating them to skills.

+In VS Code 1.129+, an experimental **prompt-file-to-skill migration tool** automates this migration. Enable it with:
+
+```json
+{
+  "chat.customizations.promptMigration.enabled": true
+}
+```
+
+Once enabled, VS Code will offer to convert your existing `*.prompt.md` files into the skill folder format automatically. Review the generated `SKILL.md` files to add `description` field improvements and any bundled assets before committing the migrated skills.
+
 **Q: Can skills include multiple files?**

 A: Yes! Skills are folders, not single files. You can bundle reference documents, templates, scripts, and any other resources the AI needs. Keep individual assets under 5 MB.