Pastewatch

Local macOS utility that obfuscates sensitive data before it is pasted into AI chat interfaces.

It operates before paste, not after submission.

If sensitive data never enters the prompt, the incident does not exist.

---

Core Principle

Principiis obsta — resist the beginnings.

Pastewatch intervenes at the earliest irreversible boundary: the moment data leaves the user's control.

Once pasted into an AI system, data cannot be reliably recalled, audited, or constrained.

Pastewatch refuses that transition.

---

What Pastewatch Does

• Monitors clipboard content locally
• Detects high-confidence sensitive data
• Obfuscates detected values before paste
• Operates fully offline
• Shows minimal, explicit feedback when changes occur

Nothing more.

---

What Pastewatch Does Not Do

Pastewatch is not:

• a DLP system
• a compliance product
• a browser extension
• an LLM proxy
• a monitoring or logging tool
• an AI-powered classifier
• a policy engine

Pastewatch does not:

• block paste
• phone home
• store clipboard history
• guess or infer
• act when uncertain

False negatives are preferred over false positives.

---

How Pastewatch Works

Pastewatch modifies clipboard text locally before it is pasted.

It scans plain text for sensitive patterns and replaces them with non-sensitive placeholders.

Pastewatch does not hide clipboard contents from the operating system or applications, and it does not provide a way to restore original values after paste.

---

Installation

From Release (Recommended)

1. Download the latest .dmg from Releases
2. Open the DMG and drag Pastewatch.app to Applications
3. Launch Pastewatch from Applications
4. Grant notification permissions when prompted

From Source

git clone https://github.com/ppiankov/pastewatch.git
cd pastewatch
swift build -c release
./.build/release/pastewatch

---

Detection Scope

Pastewatch detects only deterministic, high-confidence patterns:

Type	Examples
Email	user@company.com
Phone	+60123456789, (555) 123-4567
IP Address	192.168.1.100
AWS Keys	AKIAIOSFODNN7EXAMPLE
API Keys	sk_test_..., ghp_...
UUIDs	550e8400-e29b-41d4-a716-446655440000
JWT Tokens	eyJhbGciOiJIUzI1NiIs...
DB Connections	postgres://user:pass@host/db
SSH Keys	-----BEGIN RSA PRIVATE KEY-----
Credit Cards	4111111111111111 (Luhn validated)

No ML. No probabilistic scoring. No confidence levels.

If detection is ambiguous, Pastewatch does nothing.

---

Obfuscation Model

Detected values are replaced with stable placeholders per paste:

john.doe@example.com  →  <EMAIL_1>
AKIAIOSFODNN7EXAMPLE  →  <AWS_KEY_1>
192.168.1.100         →  <IP_1>

• Mapping exists only in memory
• Mapping is discarded immediately after paste
• No persistence
• No recovery mechanism

After paste, the system returns to rest.

---

User Experience

• Default behavior is silent

• When obfuscation occurs, a minimal notification is shown:

  Pastewatch: Obfuscated: Email (1), API Key (1)

No previews. No animations. No confirmations.

Silence is success.

---

Configuration

Optional configuration file: ~/.config/pastewatch/config.json

{
  "enabled": true,
  "enabledTypes": ["Email", "Phone", "IP", "AWS Key", "API Key", "UUID", "DB Connection", "SSH Key", "JWT", "Card"],
  "showNotifications": true,
  "soundEnabled": false
}

All settings can also be changed via the menubar dropdown.

---

Threat Model

Pastewatch assumes:

• Users will paste sensitive data
• AI systems are not trusted with raw secrets
• Prevention is cheaper than remediation

Pastewatch does not attempt to secure downstream systems. It prevents entry entirely.

---

Design Constraints

• Local-only operation
• Deterministic behavior
• Minimal UI surface
• No background analytics
• No user accounts
• No configuration required for safe defaults

If a feature increases complexity without reducing risk, it is rejected.

---

Platform Support

macOS 14+ on Apple Silicon (M1 and newer).

Intel-based Macs are not supported.

---

Project Family

Pastewatch applies Principiis obsta at the clipboard boundary. It is part of a family of tools applying the same principle at different surfaces:

Project	Boundary	Intervention Point
Chainwatch	AI agent execution	Before tool calls
Pastewatch	Data transmission	Before paste
VaultSpectre	Secrets lifecycle	Before exposure
Relay	Human connection	Before isolation compounds

Same principle. Different surfaces. Consistent philosophy.

---

Documentation

• docs/design-baseline.md — Core philosophy and design priorities
• docs/hard-constraints.md — Non-negotiable rules
• docs/status.md — Current scope and non-goals

---

License

MIT License.

Use it. Fork it. Modify it.

Do not pretend it guarantees compliance or safety.

---

Status

MVP — Experimental prototype.

The core detection and obfuscation work. Edge cases exist. Feedback welcome.