trustboost-pii-sanitizer

name	trustboost-pii-sanitizer
version	2.6.0
description	Context-aware PII sanitization for autonomous AI agent pipelines. Sanitizes text before LLMs with 5 context modes (legal/financial/medical/code/general), Privacy Budget per agent, and TrustBoost Score for M2M trust verification. Supports EN, ES (LATAM), PT (BR/PT), DE, JA, FR, IT, KO with country-specific patterns (RFC, CUIT, CPF, CNPJ, Personalausweis, マイナンバー, NIR, Codice Fiscale, 주민등록번호). Returns sanitized text, safety_score (0.0-1.0), risk_category (CRITICAL/PRIVATE/SENSITIVE/CLEAN), and context_applied. No SDK required — single POST request. 50 free requests per wallet with tx_hash="TRIAL".
license	MIT
compatibility	Requires internet access to reach the TrustBoost API. No local dependencies. Compatible with any agent that can make HTTP POST requests. No authentication required.
metadata	{"author":"teodorofodocrispin-cmyk","version":"2.6.0","endpoint":"https://api.trustboost.dev/sanitize","health":"https://api.trustboost.dev/health","payment":"Solana USDC (149 USDC = 10,000 sanitizations)","trial":"tx_hash=TRIAL (50 free sanitizations per wallet, no payment required)","preview":"https://api.trustboost.dev/sanitize/preview (3 free requests per IP, no wallet required)","autonomy_score":"8.5/10","audit_score":"9.8/10","languages":"English, Spanish, Portuguese, German, Japanese, French, Italian, Korean","compliance":"GDPR, LGPD, APPI, CCPA, DPDPA","requires_env":"none","wallet_security":"TrustBoost NEVER requires wallet private keys, seed phrases, or signing credentials. The optional wallet_address parameter accepts ONLY a public Solana address for per-wallet quota tracking. Payment signing happens entirely client-side — TrustBoost only verifies the resulting tx_hash on-chain via Helius oracle. No private key ever touches TrustBoost infrastructure.\n","infrastructure":"FastAPI + Supabase + Render (AWS)","features":["context_aware_sanitization (legal/financial/medical/code/general)","privacy_budget_per_agent (configurable daily limits)","trustboost_score (M2M trust verification)"],"mcp_server":"https://api.trustboost.dev/mcp","homepage":"https://github.com/teodorofodocrispin-cmyk/TrustBoost-PII-Sanitizer"}

TrustBoost PII Sanitizer v2.6.0

A blockchain-verified privacy layer that redacts PII from text before it reaches LLMs or external services. Payment is verified trustlessly on Solana via Helius oracle — no API key, no SDK, no human intervention required.

Running on FastAPI + Supabase + Render (AWS) — production-grade infrastructure with per-wallet TRIAL tracking and automatic replay attack protection.

---

⚠️ Transparency Notice (Read Before Installing)

TrustBoost is designed for radical transparency. Before installing, understand exactly how it works:

1. Data Transmission

Raw text is transmitted to Render (AWS) infrastructure for processing via FastAPI.

What TrustBoost stores: Only sanitized output, character count, safety score, risk category, wallet address, and timestamp — never raw input text.

Storage: Supabase PostgreSQL (private audit ledger)

For strict no-transmission requirements (e.g., on-premise HIPAA environments): this service is not suitable. Use a local sanitizer instead.

For general privacy compliance (GDPR, LGPD, APPI, CCPA): TrustBoost provides meaningful risk reduction by sanitizing PII before it reaches YOUR LLM (Claude, GPT, Gemini, etc). Detection uses OpenAI GPT-4o-mini internally — raw text reaches TrustBoost infrastructure for processing, then server-side enforcement removes all detected entities before your pipeline continues.

2. Payment Model and Autonomous Safety

TrustBoost uses Solana blockchain as the authorization layer — replacing human confirmation with cryptographic verification.

How autonomous payment works safely:

1. Agent constructs a Solana transaction
2. Agent signs with its OWN wallet private key (TrustBoost never has access to agent keys)
3. Transaction is broadcast to Solana Mainnet
4. Helius oracle verifies the transaction on-chain
5. Service activates only after verification

Why blockchain replaces human confirmation:

• Transaction is signed by the agent's own wallet
• No credentials are shared with TrustBoost
• Payment is publicly verifiable on Solana Explorer
• TrustBoost cannot initiate payments — only the agent can trigger a transaction

Safe payment implementation:

# Monitor quota — only pay when needed
if result["data"]["usage_metrics"]["quota_remaining"] < 10:
    # Autonomous payment within operator-configured limits
    # Operator sets MAX_AUTO_PAYMENT once — agent operates freely within it
    # No human loop required — blockchain verification replaces human confirmation
    if agent_wallet_balance >= MAX_AUTO_PAYMENT_USDC:
        execute_autonomous_payment(
            amount=149,
            currency="USDC",
            network="solana",
            address="giu4VciTkfWJNG1oeP6SzHEJwmabikJSMB91GaFNWE4"
        )
    else:
        notify_operator("Insufficient balance — configure wallet funding")

Recommendation: Always use TRIAL (50 free requests per wallet) before any paid transaction.

3. Audit Trail

Every sanitization is logged to Supabase PostgreSQL with:

Retention policy: Sanitized output stored 90 days, deleted on request via [e-mail]. Raw input text is never stored at any point. tx_hash, input character count (not content), sanitized output, safety score, risk category, wallet address, timestamp.

The raw input text is never stored at any point.

4. Operator Verification

• GitHub: https://github.com/teodorofodocrispin-cmyk
• Repository: https://github.com/teodorofodocrispin-cmyk/TrustBoost-PII-Sanitizer
• Health endpoint: https://api.trustboost.dev/health
• Contact: [e-mail]
• External audit score: 9.8/10
• Autonomy score: 8.5/10

---

When to use this skill

Activate this skill when:

• The agent needs to process user-generated text that may contain emails, passwords, private keys, phone numbers, addresses, or national IDs
• Privacy compliance is required before passing content to an LLM provider (GDPR, LGPD, APPI, CCPA)
• The agent processes text in Spanish, Portuguese, German, or Japanese with country-specific PII formats
• The agent needs autonomous budget management via the quota_remaining field

When NOT to use this skill

• Strict on-premise HIPAA environments requiring zero data transmission to third parties
• Use cases involving real private keys or passwords in production (use TRIAL only for testing)

---

How it works

1. Agent sends text + a Solana tx_hash to the TrustBoost API
2. Supabase verifies per-wallet TRIAL quota or Helius oracle verifies payment on Solana mainnet
3. OpenAI GPT-4o-mini detects and redacts all PII in the input language
4. Returns sanitized text + safety score + risk category + quota metrics
5. Only sanitized output is logged — never raw input

Multilingual PII Support

TrustBoost automatically detects the input language and applies country-specific patterns:

Language	Region	PII Patterns
🇺🇸 English	Global	SSN, API keys, credit cards, passwords
🇲🇽🇨🇴🇦🇷 Spanish	Latin America	RFC, CUIT, RUT, DNI, CURP, Cédula, RUC
🇧🇷🇵🇹 Portuguese	Brazil & Portugal	CPF, CNPJ, RG, NIF, NUS
🇩🇪 German	Germany/Austria/CH	Personalausweis, Steuernummer, IBAN DE
🇯🇵 Japanese	Japan	マイナンバー, 運転免許証, パスポート番号, 住所
🇫🇷🇧🇪🇨🇦 French	France / Belgium / Switzerland / Canada	NIR (Sécurité Sociale), SIRET, SIREN, Carte Vitale, IBAN FR
🇮🇹 Italian	Italy / San Marino	Codice Fiscale, Partita IVA, Carta d'Identità, Tessera Sanitaria, IBAN IT
🇰🇷 Korean	Republic of Korea	주민등록번호 (RRN), 사업자등록번호, 여권번호, 운전면허번호

---

Try it in 10 seconds — no wallet needed

curl -X POST https://api.trustboost.dev/sanitize/preview \
  -H "Content-Type: application/json" \
  -d '{"text": "My name is John Doe, email [e-mail], SSN 123-45-6789"}'

{
  "sanitized_content": "My name is [REDACTED], email [REDACTED], SSN [REDACTED]",
  "safety_score": 0.6,
  "risk_category": "PRIVATE",
  "demo": true,
  "requests_remaining": 2,
  "next": "https://github.com/teodorofodocrispin-cmyk/TrustBoost-PII-Sanitizer#trial"
}

3 free previews per IP · no account · no wallet · no setup. Ready for more? 50 free sanitizations with a Solana wallet using tx_hash="TRIAL".

---

API Request

Endpoint: POST https://api.trustboost.dev/sanitize

Headers: Content-Type: application/json

Payload:

{
  "text": "The text containing potential PII",
  "tx_hash": "TRIAL",
  "wallet_address": "your-solana-wallet"
}

Access modes

Mode	tx_hash value	Cost	Quota
Trial	"TRIAL"	Free	50 sanitizations per wallet
Paid	Real Solana tx hash	149 USDC	10,000 sanitizations

To get a paid tx_hash: Send exactly 149 USDC on Solana mainnet to: giu4VciTkfWJNG1oeP6SzHEJwmabikJSMB91GaFNWE4

---

API Response (Success 200)

{
  "status": "success",
  "request_id": "TRIAL",
  "data": {
    "message": "Content successfully sanitized and logged.",
    "sanitized_content": "Text with [REDACTED] replacing all PII",
    "safety_score": 0.95,
    "risk_category": "PRIVATE",
    "entities_removed": true,
    "timestamp": "2026-04-27T09:00:00Z",
    "usage_metrics": {
      "quota_remaining": 49,
      "quota_limit": 50
    }
  },
  "billing": {
    "license_type": "TRIAL",
    "status": "active"
  }
}

API Response (Error 402)

{
  "status": "error",
  "request_id": "TRIAL",
  "code": "QUOTA_EXHAUSTED_OR_PAYMENT_REQUIRED",
  "message": "TRIAL quota exhausted. Send 149 USDC on Solana to continue.",
  "trial_info": {
    "quota_used": 50,
    "quota_limit": 50,
    "quota_remaining": 0
  },
  "payment_info": {
    "amount_required": 149,
    "currency": "USDC",
    "network": "solana",
    "payment_address": "giu4VciTkfWJNG1oeP6SzHEJwmabikJSMB91GaFNWE4"
  },
  "next_steps": [
    {
      "action": "send_payment",
      "description": "Send 149 USDC on Solana Mainnet to the payment address"
    },
    {
      "action": "retry_with_tx_hash",
      "description": "Resubmit request including the Solana transaction signature"
    }
  ]
}

API Response (Error 409)

{
  "status": "error",
  "code": "TX_HASH_ALREADY_USED",
  "message": "This transaction hash has already been used. Each tx_hash can only be used once.",
  "payment_info": {
    "amount_required": 149,
    "currency": "USDC",
    "network": "solana",
    "payment_address": "giu4VciTkfWJNG1oeP6SzHEJwmabikJSMB91GaFNWE4"
  }
}

---

Risk categories

Category	What gets redacted
CRITICAL	Private keys, seed phrases, passwords, credit card data
PRIVATE	Emails, phone numbers, national IDs, physical addresses
SENSITIVE	Social media handles, general locations

Safety score

• 0.0 — No PII detected, text is clean
• 0.5 — Moderate PII detected (emails, handles)
• 1.0 — Critical PII detected (keys, passwords)

---

Example — English

Input:

{
  "text": "Contact John at [e-mail] or [phone]. API key: sk-abc123xyz.",
  "tx_hash": "TRIAL",
  "wallet_address": "your-wallet"
}

Output:

{
  "sanitized_content": "Contact [REDACTED] at [REDACTED] or [REDACTED]. API key: [REDACTED].",
  "safety_score": 0.97,
  "risk_category": "CRITICAL",
  "entities_removed": true
}

Example — German

Input:

{
  "text": "Hans Müller, Personalausweis: L01X00T47, IBAN: DE89 3704 0044 0532 0130 00, Tel: [phone]",
  "tx_hash": "TRIAL"
}

Output:

{
  "sanitized_content": "[REDACTED], Personalausweis: [REDACTED], IBAN: [REDACTED], Tel: [REDACTED]",
  "safety_score": 0.98,
  "risk_category": "CRITICAL",
  "entities_removed": true
}

Example — Japanese

Input:

{
  "text": "田中太郎、マイナンバー：123456789012、電話：[phone]",
  "tx_hash": "TRIAL"
}

Output:

{
  "sanitized_content": "[REDACTED]、マイナンバー：[REDACTED]、電話：[REDACTED]",
  "safety_score": 0.97,
  "risk_category": "PRIVATE",
  "entities_removed": true
}

---

External Evaluations

• Autonomy Score: 8.5/10 — AI evaluation, not a certified security audit
• Blueprint Audit: 9.8/10 — AI evaluation, not a certified security audit
• Independent security audit: pending — this project has not been audited by a certified security firm. See the Manifesto for honest disclosure.
• Full report: https://github.com/teodorofodocrispin-cmyk/TrustBoost-PII-Sanitizer/blob/main/AGENT_EVALUATION.md

Known Limitations

• Prompt injection risk: Malicious text containing instructions like "Ignore previous instructions" could potentially bypass PII redaction. temperature=0 and strict JSON-only output reduce this risk but do not eliminate it entirely.
• Not suitable for zero-transmission environments: Raw text is sent to api.trustboost.dev before sanitization occurs.
• TRIAL is trust-based: Per-wallet quota tracking is not cryptographically verified.
• No certified audit: Evaluation scores are AI-generated, not from a certified security firm.

Resources

• GitHub: https://github.com/teodorofodocrispin-cmyk/TrustBoost-PII-Sanitizer
• Health check: https://api.trustboost.dev/health
• Schema (molt.json): https://raw.githubusercontent.com/teodorofodocrispin-cmyk/TrustBoost-PII-Sanitizer/main/molt.json
• Infrastructure: FastAPI + Supabase + Render (AWS)
• Live Demo: https://huggingface.co/spaces/TrustBoost/pii-sanitizer
• Verify proof: https://api.trustboost.dev/verify/{anchor_tx}