Code Instructor
A VS Code / Cursor extension for AI-assisted coding: real-time development progress, goal deviation tracking, and code validity checks.
Features
- Sidebar Dashboard — Progress signals, plan steps, chat assistant, and event timeline
- Goal & Plan Baseline — Import from
goal.md / plan.md or clipboard
- Deviation Score — Heuristic + optional LLM-based assessment
- Build Validity Check — Configurable build/test commands
- Code Instructor Chat — Q&A about current goal, plan, and progress
Installation
From VS Code Marketplace
- Open VS Code or Cursor
- Press
Ctrl+Shift+X (Windows/Linux) or Cmd+Shift+X (macOS) to open Extensions
- Search for Code Instructor
- Click Install
From VSIX (local)
- Download the
.vsix file from Releases
- Press
Ctrl+Shift+P / Cmd+Shift+P → Extensions: Install from VSIX...
- Select the downloaded
.vsix file
Quick Start
Click the Code Instructor icon in the left activity bar, or run the command Code Instructor: Open Sidebar.
2. Import goal and plan
Option A — From workspace files (recommended)
Create two files in your workspace root:
goal.md — Project goal with # Title and ## User block for summaryplan.md — Execution plan with checkboxes (- [ ] / - [x])
The extension auto-imports on startup. Or click Import goal to import manually.
Option B — From clipboard
- Copy goal/plan text from a conversation or document
- Click Import goal in the sidebar
- If clipboard content is invalid, a manual input box will appear
3. View progress
After importing, the sidebar shows:
- Status — Active / Lagging / Blocked
- Progress — Plan completion (e.g. 25/26)
- Deviation — Low / Medium / High
- Validity — Build check state (idle / running / passed / failed)
4. Run build check
Click Run Check in the Validity section to run your build command. Configure aiDevCoach.validity.buildCommand if needed (default: auto-infer from package.json scripts).
5. Use Code Instructor chat
Type a question in the chat input and press Enter. The assistant answers based on current goal, plan, and evidence. Cloud LLM is optional (see Settings).
Settings
| Setting |
Description |
Default |
aiDevCoach.validity.buildCommand |
Build command for validity check |
Auto-infer from scripts |
aiDevCoach.validity.testCommand |
Optional test command |
Auto-infer |
aiDevCoach.validity.runTestAfterBuild |
Run test after build passes |
false |
aiDevCoach.validity.cwd |
Working directory for commands |
Workspace root |
aiDevCoach.cloud.enabled |
Enable cloud LLM for assessment/chat |
false |
aiDevCoach.cloud.provider |
Provider: deepseek, openai, anthropic, custom |
deepseek |
aiDevCoach.cloud.apiKey |
API key for the provider |
(empty) |
aiDevCoach.ingest.planFilePath |
Custom plan file path |
(empty) |
Note: Cloud calls are optional. If disabled, chat uses local heuristics based on current signals.
Commands
| Command |
Description |
| Code Instructor: Open Sidebar |
Open the Code Instructor sidebar |
| Code Instructor: Import Baseline From goal.md/plan.md |
Import goal and plan from workspace files |
| Code Instructor: Set Goal Baseline From Clipboard |
Import goal from clipboard (fallback when files missing) |
| Code Instructor: Import Plan Steps From Clipboard |
Import plan steps from clipboard |
| Code Instructor: Run Validity Check |
Run build/test check |
| Code Instructor: Add Plan Step |
Add a plan step manually |
Development
npm install
npm run build
Press F5 to launch the Extension Development Host. Open the sidebar via the Code Instructor icon.
Run tests
npm run test
License
MIT License. See LICENSE for details.
