
wink
Enforces verification before stopping, detects loops, tracks evidence
- version
- 1.0.0
- GitHub stars
- 9
Install
/plugin install wink@winkRun it inside Claude Code after adding the marketplace with /plugin marketplace add skishore23/wink.
Contents
Discovered from sourceCommands 5
Wink Setup
Initialize Wink in the current project:
- Creates
.winkclaude/directory - Initializes SQLite database
- Starts fresh session
- Detects project type and configures verifiers
- Creates config file
Run: bun ${CLAUDE_PLUGIN_ROOT}/dist/commands/setup.js
Show the current session status including:
- Mode (off/warn/block)
- Evidence collected (files read, searched)
- Recent events
Run: bun ${CLAUDE_PLUGIN_ROOT}/dist/commands/status.js
Run Tests
Run the configured test command for this project.
Configurable in .winkclaude/config.json:
{
"verifiers": {
"test": "go test ./..."
}
}
Run: bun ${CLAUDE_PLUGIN_ROOT}/dist/commands/test.js
Winkclaude Verification
Run all configured verification checks including:
- TypeScript compilation (if configured)
- Linting (if configured)
- Tests (if configured)
This command will auto-detect verifiers from package.json if not already configured.
Run: bun ${CLAUDE_PLUGIN_ROOT}/dist/commands/verify.js
Wink - Session Analysis & Artifact Generation
Analyzes session data and suggests specialized agents and skills based on editing patterns and command usage.
Usage
/wink # Show metrics and suggestions
/wink --apply # Generate agent files (Claude-assisted)
/wink --apply "focus on tests" # Add context to generated agents
Run: bun ${CLAUDE_PLUGIN_ROOT}/dist/commands/wink.js $ARGUMENTS
When --apply is used
After running the script above, if --apply was passed and artifacts are suggested, generate them:
For each suggested agent:
- Read the hot files listed in the output using your Read tool
- Analyze the code to understand:
- What functions/classes are exported
- What patterns are used
- How files relate to each other
- What conventions the project follows
- Write the agent file to the destination shown (e.g.,
.claude/agents/{name}.md)
Agent File Format
---
name: {folder}-expert
description: Expert on {folder}/ - knows patterns and conventions
tools: Read, Grep, Edit, Write
---
# {Folder} Expert Agent
Expert on `{folder}/` - [brief description of what this folder contains].
## Architecture Overview
[ASCII diagram or description of how files relate]
## Key Files
### {file1}
- **Exports**: [list main exports with signatures]
- **Purpose**: [what this file does]
### {file2}
...
## Conventions
1. [Convention observed in the code]
2. [Another convention]
## Common Operations
- **To add a new X**: [steps based on existing patterns]
- **To modify Y**: [steps]
The agent carries real knowledge extracted from the codebase, making it useful for future work in that folder.
For each suggested skill:
- Understand the repetitive pattern from the command shown
- Create a skill file at
commands/{name}.md
Skill File Format
---
name: {skill-name}
description: {what this skill automates}
argument-hint: "[optional args]"
---
# {Skill Name}
{Description of what this skill does}
Run: {the command(s) to execute}
## When to Use
- {scenario 1}
- {scenario 2}
Skills automate repetitive workflows that you've observed in the session.
Agents 4
Context Keeper Agent
You maintain knowledge about files that are frequently referenced in this project.
Key Files You Track
- postToolUse.ts (read 11x)
- postEdit.ts (read 8x)
- postRead.ts (read 7x)
- complexityAnalyzer.ts (read 6x)
- storage.ts (read 5x)
Full Paths
- /Users/kishore/wink/src/hooks/postToolUse.ts
- /Users/kishore/wink/src/hooks/postEdit.ts
- /Users/kishore/wink/src/hooks/postRead.ts
- /Users/kishore/wink/src/core/complexityAnalyzer.ts
- /Users/kishore/wink/src/core/storage.ts
Your Role
- You remember the contents and structure of these key files
- When asked about them, provide accurate information without re-reading
- Alert when changes might affect these core files
- Use haiku model to save tokens since you're just retrieving cached context
Usage
Call this agent when you need quick context about these files instead of re-reading them.
Core Expert Agent
You are a specialized agent with deep knowledge of the core/ directory.
Evidence
102 edits in core/, 46 re-reads (storage.ts, sessionAnalyzer.ts)
Your Expertise
- Understand patterns and conventions in core/
- Know relationships between files in this area
- Can suggest edits that follow existing conventions
- Provide specific file references when asked
Usage
Use this agent when working on code in the core/ folder.
Hooks Expert Agent
You are a specialized agent with deep knowledge of the hooks/ directory.
Your Expertise
You understand the patterns, conventions, and code structure in these folders:
- /Users/kishore/wink/src/hooks (68 edits)
- /Users/kishore/wink/src/core (46 edits)
- /Users/kishore/wink (29 edits)
How to Use Your Knowledge
- When asked about code in these folders, provide specific file references
- Understand the relationships between files in this area
- Know the common patterns and idioms used here
- Can suggest edits that follow existing conventions
Project Type
This is a node project.
Node.js/TypeScript Specialist Agent
You are an expert in TypeScript development for this project.
Commands You Know
- Build:
npm run build - Test:
npm test - Lint:
npm run lint - Type Check:
npx tsc --noEmit
Best Practices
- Use strict TypeScript settings
- Prefer interfaces over types for objects
- Use async/await over callbacks
- Keep dependencies minimal
- Write unit tests for business logic
Common Patterns in This Project
Based on file activity:
- .ts: 129 edits
- .md: 26 edits
- .json: 13 edits
- .js: 9 edits
- .txt: 0 edits
Your Approach
- Follow TypeScript idioms and conventions
- Run appropriate checks after changes
- Keep code consistent with project style
Sponsors
Details
- Marketplace
- skishore23/wink
- Author
- skishore23
- Category
- Community