feat(plugins): Add secrets-scanner plugin for credential exposure detection

plugins/secrets-scanner/.claude-plugin/plugin.json

@@ -0,0 +1,9 @@
+{
+  "name": "secrets-scanner",
+  "version": "1.0.0",
+  "description": "Real-time detection of exposed secrets, API keys, and credentials in Claude Code output. Warns users when sensitive data is detected to prevent accidental exposure.",
+  "author": {
+    "name": "Steven Elliott",
+    "email": "srichardelliottjr@gmail.com"
+  }
+}

plugins/secrets-scanner/README.md

@@ -0,0 +1,231 @@
+# Secrets Scanner Plugin for Claude Code
+
+Real-time detection of exposed secrets, API keys, and credentials in Claude Code output. Warns users immediately when sensitive data is detected to prevent accidental exposure.
+
+## Why This Plugin?
+
+Claude Code can read files, execute commands, and display their output. When working with configuration files, environment variables, or logs, sensitive credentials may be inadvertently exposed in:
+
+- Terminal output (visible to anyone watching)
+- Terminal history (persisted in shell history files)
+- Claude's cached tool results (`~/.claude/projects/*/tool-results/`)
+- Screen recordings or screenshots
+
+This plugin provides defense-in-depth by scanning all tool output for secret patterns and alerting you immediately.
+
+**Addresses Issues:**
+- [#18223](https://github.com/anthropics/claude-code/issues/18223) - Claude displayed sensitive .env credentials
+- [#22548](https://github.com/anthropics/claude-code/issues/22548) - Auto-mask API keys in VS Code extension
+- [#21528](https://github.com/anthropics/claude-code/issues/21528) - Environment Variable Redaction
+
+## Features
+
+### Comprehensive Secret Detection
+
+Detects 35+ types of secrets including:
+
+| Category | Examples |
+|----------|----------|
+| **AI Services** | Anthropic, OpenAI API keys |
+| **Cloud Providers** | AWS, Google Cloud, Azure credentials |
+| **Code Platforms** | GitHub tokens, npm tokens |
+| **Payment** | Stripe live/test keys |
+| **Communication** | Slack, Twilio, SendGrid tokens |
+| **Databases** | PostgreSQL, MySQL, MongoDB, Redis connection strings |
+| **Cryptographic** | RSA, SSH, PGP private keys |
+| **Auth Tokens** | JWTs, Bearer tokens, Basic auth |
+
+### Severity Levels
+
+- 🚨 **Critical**: Production API keys, private keys, database credentials
+- ⚠️ **High**: OAuth tokens, webhook URLs, service tokens
+- ⚡ **Medium**: Test keys, generic API keys
+- ℹ️ **Low**: Potentially sensitive patterns
+
+### Security Logging
+
+All detections are logged (without actual secret values) to:
+```
+~/.claude/security-logs/secrets-detected-YYYY-MM-DD.jsonl
+```
+
+This provides an audit trail for security review without persisting the secrets themselves.
+
+## Installation
+
+### Option 1: Copy to plugins directory
+
+```bash
+cp -r secrets-scanner ~/.claude/plugins/
+```
+
+### Option 2: Add to settings.json
+
+```json
+{
+  "plugins": [
+    "/path/to/secrets-scanner"
+  ]
+}
+```
+
+## Usage
+
+Once installed, the plugin automatically scans all tool output. No configuration required.
+
+### Example Warning
+
+When secrets are detected, you'll see:
+
+```
+============================================================
+🔐 SECRETS DETECTED IN OUTPUT
+============================================================
+
+🚨 CRITICAL:
+   • Stripe Secret Key - Stripe live secret key
+   • AWS Access Key ID - AWS access key ID
+
+⚠️  HIGH:
+   • GitHub Token (Classic) - GitHub personal access token
+
+Source: .env
+
+⚡ RECOMMENDED ACTIONS:
+   1. Do NOT share this terminal output
+   2. Clear terminal history: `history -c`
+   3. Rotate exposed credentials immediately
+   4. Check ~/.claude/projects/ for cached data
+
+📋 This detection has been logged to:
+   ~/.claude/security-logs/
+
+============================================================
+```
+
+### Reviewing Security Logs
+
+```bash
+# View today's detections
+cat ~/.claude/security-logs/secrets-detected-$(date +%Y-%m-%d).jsonl | jq '.'
+
+# Count detections by type
+cat ~/.claude/security-logs/secrets-detected-*.jsonl | jq -r '.detections[].type' | sort | uniq -c
+
+# Find critical detections
+cat ~/.claude/security-logs/secrets-detected-*.jsonl | jq 'select(.detections[].severity == "critical")'
+```
+
+## Secret Patterns
+
+### AI Service Keys
+
+| Pattern | Example |
+|---------|---------|
+| Anthropic API Key | `sk-ant-api03-...` |
+| OpenAI API Key | `sk-...` (48 chars) |
+| OpenAI Project Key | `sk-proj-...` |
+
+### Cloud Providers
+
+| Pattern | Example |
+|---------|---------|
+| AWS Access Key ID | `AKIA...` (20 chars) |
+| AWS Secret Key | Via `aws_secret_access_key=` |
+| Google API Key | `AIza...` (39 chars) |
+| Azure Storage Key | Connection string pattern |
+
+### Code Platforms
+
+| Pattern | Example |
+|---------|---------|
+| GitHub Classic Token | `ghp_...` (40 chars) |
+| GitHub Fine-grained | `github_pat_...` |
+| npm Token | `npm_...` (40 chars) |
+
+### Database Connection Strings
+
+Detects credentials embedded in connection strings:
+- `postgres://user:password@host/db`
+- `mysql://user:password@host/db`
+- `mongodb+srv://user:password@cluster/db`
+- `redis://user:password@host:port`
+
+### Private Keys
+
+Detects key file headers:
+- `-----BEGIN RSA PRIVATE KEY-----`
+- `-----BEGIN OPENSSH PRIVATE KEY-----`
+- `-----BEGIN PGP PRIVATE KEY BLOCK-----`
+- `-----BEGIN EC PRIVATE KEY-----`
+
+## Configuration
+
+Currently, the plugin uses built-in patterns. Future versions may support:
+
+```json
+{
+  "secretsScanner": {
+    "additionalPatterns": [...],
+    "excludePatterns": [...],
+    "minSecretLength": 20,
+    "logDetections": true
+  }
+}
+```
+
+## Best Practices
+
+### When Secrets Are Detected
+
+1. **Don't panic** - The warning is to help you, not punish you
+2. **Don't share** - Avoid copying/pasting terminal output
+3. **Clear history** - Run `history -c` and clear scrollback
+4. **Rotate credentials** - Assume they may be compromised
+5. **Check caches** - Look in `~/.claude/projects/` for persisted data
+
+### Preventive Measures
+
+1. **Use secret managers** - Don't store secrets in plain text files
+2. **Use .gitignore** - Prevent .env files from being committed
+3. **Environment variables** - Load secrets from env, not files
+4. **Least privilege** - Use API keys with minimal permissions
+
+## Limitations
+
+- **Detection only** - This plugin warns but cannot prevent Claude from seeing the content
+- **Pattern-based** - Novel secret formats may not be detected
+- **Post-execution** - Scanning happens after the tool runs
+- **No redaction** - Full content is still displayed (redaction requires core changes)
+
+For true prevention, consider:
+- Using environment variables instead of files
+- Implementing a proxy that redacts before display
+- Waiting for core redaction support ([#18653](https://github.com/anthropics/claude-code/issues/18653))
+
+## Privacy
+
+This plugin:
+- ✅ Logs detection metadata (type, severity, source file)
+- ❌ Does NOT log actual secret values
+- ❌ Does NOT send data externally
+- ✅ All logs stay local in `~/.claude/security-logs/`
+
+## Contributing
+
+Contributions welcome:
+- Additional secret patterns
+- Improved detection accuracy
+- Reduced false positives
+- Configuration options
+
+## License
+
+MIT License
+
+## Author
+
+Steven Elliott - CIO with expertise in compliance (HIPAA, SOC 2, ERISA) and AI transformation in regulated environments.
+
+- GitHub: [@stevenelliottjr](https://github.com/stevenelliottjr)
+- LinkedIn: [/in/stevenelliottjr](https://linkedin.com/in/stevenelliottjr)

plugins/secrets-scanner/hooks/hooks.json

@@ -0,0 +1,16 @@
+{
+  "description": "Secrets scanner - detects exposed credentials in tool output",
+  "hooks": {
+    "PostToolUse": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "python3 ${CLAUDE_PLUGIN_ROOT}/hooks/posttooluse.py",
+            "timeout": 10
+          }
+        ]
+      }
+    ]
+  }
+}