feat: add @aicoach chat participant for AI Engineer Coach data access

README.extension.md

@@ -61,12 +61,17 @@ The extension is organized into three sections: **Observe**, **Measure**, and **
 | **OpenCode** | macOS/Linux: `~/.local/share/opencode/`<br>Windows: `%USERPROFILE%\.local\share\opencode\` |
 | **GitHub Copilot CLI** | `~/.copilot/session-state/` and `~/.copilot/history-session-state/` |
 
+### Chat
+
+Type `@aicoach` in any VS Code chat panel for conversational access to all coaching data. Slash commands `/summary`, `/improve`, `/compare`, and `/flow` give quick access to common analyses. The participant orchestrates multiple backend tools automatically to answer complex questions.
+
 ## Getting Started
 
 1. Open the command palette (`Cmd+Shift+P` / `Ctrl+Shift+P`).
 2. Run **AI Engineer Coach: Open Dashboard**.
 3. Use the sidebar to navigate pages. Filter by workspace or harness at the bottom.
 4. Run **AI Engineer Coach: Reload Data** to re-parse after new sessions.
+5. Type `@aicoach` in VS Code chat for conversational coaching.
 
 
 

cspell.json

@@ -3,6 +3,7 @@
   "language": "en",
   "minWordLength": 5,
   "words": [
+    "aicoach",
     "affordances",
     "allpending",
     "antipatterns",

docs/content/features/_index.md

@@ -26,6 +26,10 @@ AI Engineer Coach organizes its capabilities into three areas that mirror a cont
 - [Skill Finder](/improve/skill-finder/) -- Discover repeated prompts and matching community skills
 - [Context Health](/improve/context-health/) -- Evaluate context quality and session management efficiency
 
+## Chat
+
+- [Chat Participant](/features/chat/) -- Conversational access to all coaching data via `@aicoach` in VS Code chat, with slash commands and agentic tool orchestration
+
 ## Level Up
 
 - [Learning Center](/level-up/learning/) -- Personalized quizzes and challenges built from your actual usage

docs/content/features/chat.md

@@ -0,0 +1,95 @@
+---
+title: "Chat Participant"
+weight: 30
+description: "Conversational access to all coaching data via @aicoach in VS Code chat"
+---
+
+# Chat Participant
+
+The `@aicoach` chat participant gives you conversational access to all AI Engineer Coach data directly in the VS Code chat panel. Ask questions in natural language and get data-driven coaching responses without leaving your editor.
+
+## Getting Started
+
+Before using `@aicoach`, open the AI Engineer Coach sidebar at least once to load your session data.
+
+> **Note:** When you use `@aicoach`, your question and the results returned by AI Engineer Coach tools are sent to the selected VS Code chat model so it can synthesize a response. The underlying coaching data is gathered locally, but the final chat answer is not produced purely through local processing.
+
+Type `@aicoach` in any VS Code chat panel followed by your question:
+
+```
+@aicoach how am I doing this week?
+```
+
+The participant is sticky — once invoked, follow-up messages in the same thread continue the conversation without needing to type `@aicoach` again.
+
+## Slash Commands
+
+| Command | Description | Default prompt |
+|---|---|---|
+| `/summary` | Quick usage overview | Highlights strengths and top areas to improve |
+| `/improve` | Improvement recommendations | Top 3 things to improve with specific actions |
+| `/compare` | Tool comparison | Compare AI coding tools and their effectiveness |
+| `/flow` | Flow & focus analysis | Deep work patterns and best productivity hours |
+
+Use a slash command with no additional text to get the default analysis, or add your own question:
+
+```
+@aicoach /flow Am I more productive in the morning or afternoon?
+```
+
+## Available Tools
+
+The participant has access to 12 backend tools that it selects automatically based on your question:
+
+| Tool | Domain | What it returns |
+|---|---|---|
+| `aiEngineerCoach_summary` | Observe | Session counts, recommendations, top anti-patterns |
+| `aiEngineerCoach_activity` | Observe | Daily requests, LOC, sessions, and harness breakdown |
+| `aiEngineerCoach_credits` | Measure | AI credit usage, per-model breakdown, daily trend, and costly requests |
+| `aiEngineerCoach_codeProduction` | Measure | AI vs user LOC, language breakdown, workspace distribution |
+| `aiEngineerCoach_flow` | Measure | Deep work scores, best hours, follow-up latency |
+| `aiEngineerCoach_patterns` | Improve | Anti-patterns and practice recommendations with severity |
+| `aiEngineerCoach_insights` | Improve | Learning velocity, intent classification, prompt maturity |
+| `aiEngineerCoach_wellbeing` | Improve | Work-life balance score, time distribution, burnout risk |
+| `aiEngineerCoach_workflows` | Improve | Repeated workflow clusters with automation suggestions |
+| `aiEngineerCoach_harnessComparison` | Observe | Side-by-side tool comparison: sessions, LOC, cancel rates |
+| `aiEngineerCoach_sessions` | Observe | Browse or search individual sessions by ID or keyword |
+| `aiEngineerCoach_contextHealth` | Improve | Context utilization, compaction, config health, and instruction quality |
+
+All tools accept optional `fromDate`, `toDate`, `workspaceId`, and `harness` filters. The participant resolves relative time references ("last week", "past month") automatically.
+
+## How It Works
+
+The participant runs an **agentic loop** that:
+
+1. Sends your question along with a coaching persona and tool-selection heuristics to the language model
+2. The model decides which tools to call based on your intent
+3. Tool results are fed back into the conversation for the model to synthesize
+4. The model may call additional tools if needed (up to 8 rounds)
+5. A final, synthesized coaching response is streamed back to you
+
+This means a single question like "compare my productivity this week vs last week" can trigger multiple tool calls (activity, flow, code production) and produce a unified answer.
+
+## Example Conversations
+
+**Broad check-in:**
+```
+@aicoach Give me a quick health check
+```
+→ Calls `summary`, returns practice scores, session count, top anti-pattern, and a suggested next step.
+
+**Specific investigation:**
+```
+@aicoach Why is my prompt quality score dropping?
+```
+→ Calls `patterns` with recent date range, surfaces the specific anti-patterns driving the score down with example prompts from your sessions.
+
+## Follow-ups
+
+After each response, the participant suggests follow-up prompts to guide deeper analysis:
+
+- **Improve** — "What should I improve next?"
+- **Compare tools** — "Compare my AI tools"
+- **Flow state** — "How is my focus & flow?"
+
+Click any follow-up to continue the conversation without typing.