plato-input-sanitizer

Input validation and authentication middleware for PLATO fleet services.

Components

input_sanitizer — Defense-in-depth input validation

• Agent name sanitization (XSS, SQL injection, path traversal blocking)
• Room name validation (whitelist characters)
• Content sanitization (null bytes, length limits)
• Tile submission validation (required fields, answer min-length)

auth — Bearer token authentication with rate limiting

• Simple shared-secret Bearer token auth
• Per-agent rate limiting (configurable RPM for reads vs writes)
• Timing-safe token comparison
• Modes: token (enforce), open (dev), permissive (log-only)

Install

pip install plato-input-sanitizer

Usage

from plato_input_sanitizer import sanitize_agent_name, sanitize_tile_submission
from plato_input_sanitizer.auth import AuthMiddleware

# Validate inputs
name = sanitize_agent_name("agent-42")  # raises ValueError if bad

# Auth middleware
auth = AuthMiddleware.from_env()  # reads PLATO_API_KEY
if not auth.validate(request.headers.get("Authorization", "")):
    return 401, {"error": "Unauthorized"}

if not auth.check_rate_limit(agent_name, is_write=True):
    return 429, {"error": "Rate limited"}

Security Audit Response

Created in response to external security audit findings:

• CRITICAL-5: Stored XSS via agent names → sanitize_agent_name() with HTML escaping
• HIGH-4: Path traversal in multiple services → injection pattern blocking
• HIGH-3: Rate limit bypass via agent rotation → per-agent rate limiting with timing enforcement