Add 402Sentinel skill — pre-payment counterparty risk + compliance gate for agents (x402)

[kaditang]

Adds the 402Sentinel skill (402sentinel/) — a pre-payment safety + compliance gate for Bankr agents. Call it before paying an x402 counterparty. No signup, no API key; every paid endpoint speaks x402 (USDC on Base eip155:8453 or Solana).

What an agent gets

• Seller risk (/api/assess, $0.002) — score a counterparty before paying: risk_score 0-100 + allow/review/block + key_factors (the signals that drove it, so the decision is legible) + a wallet-ready recommended_policy. EVM (Base) or Solana sellers. /api/assess/deep ($0.02) scans more history.
• Buyer-side payment firewall (/api/firewall, $0.002) — vets THIS payment against the agent's own behaviour: routing/amount/velocity anomalies, injection_destination (pass the untrusted page/tool output — if the payTo appears in it, the destination was injected → hard block), and intent_mismatch (pass the intended payTo/max — any deviation → hard block). Stateful per agent_id, with signed anti-poison history.
• Compliance (/api/compliance, $0.02) — OFAC sanctions (hit = block), due-diligence tier, structuring detection, FATF Travel-Rule (informational). Screening support, not legal advice.
• Tokenized-RWA gate (/api/assess/rwa, $0.02) — "is this TOKEN what it claims?" proxy-aware transfer-restriction detection (ERC-1404/ERC-3643/allowlist/freeze/pause), behavioral transfer simulation (reverts = trapped funds), live pause state, issuer screening.
• Verification (/api/verify, $0.02, opt-in) — a positive-only signed trust credential + public allow-list.
• Free: /api/assess/preview (verdict-only, confirm the gate before paying), /api/verified + /api/verification (public allow-list / credential), /api/report_outcome.

Skill contents

• catalog.json — schemaVersion 1, x402 demo (@x402/fetch + ExactEvmScheme, POST bodies), Bankr-wallet-signing preferred.
• skill.md — access model, x402 call checklist, deterministic preflight, and the full priced endpoint catalog.
• README row added to the Available Skills table.

Self-describes zero-config at https://402sentinel.com/.well-known/x402 — algorithmic risk signal, informational only (not advice/endorsement/accusation).

🤖 Generated with Claude Code