promptshield-llm 0.1.0

Redact sensitive data from LLM prompts before sending them to a model.

promptshield-llm

Redact sensitive data from LLM prompts before sending them to a model.

Installation

pip install promptshield-llm

Usage

from promptshield import shield_prompt

prompt = "Summarize this email from [e-mail]. Token: Bearer abc123456789"
safe_prompt = shield_prompt(prompt)

print(safe_prompt)

Output

Summarize this email from [EMAIL]. Token: [TOKEN]

Find sensitive data

from promptshield import find_sensitive

matches = find_sensitive("Contact me at [e-mail]")
print(matches)

Custom patterns

from promptshield import shield_prompt

safe = shield_prompt(
    "Customer ID: CUST-12345",
    custom_patterns={"customer_id": r"CUST-\d+"}
)

print(safe)

Overview

promptshield-llm is a tiny Python utility for masking sensitive values in prompts, logs, and LLM inputs.

It is useful when building:

• LLM applications
• RAG pipelines
• AI agents
• prompt logging systems
• internal AI tools

Features

• Redacts emails
• Redacts phone numbers
• Redacts API keys and tokens
• Redacts sensitive URLs
• Supports custom regex patterns
• Uses the Python standard library
• Simple API

Limitations

promptshield-llm is regex-based and may not catch every possible secret or personal identifier. Use it as an extra safety layer, not as your only security control.

Issues

Report issues at: https://github.com/edujbarrios/promptshield-llm

Author

Eduardo J. Barrios [e-mail]

License

Mozilla Public License 2.0