mind-intelligence-skill

Two complementary layers for autonomous Solana agents: intelligence (verify before acting) + x402 payments (pay to access, prove what happened).

A skill for the Solana AI Kit.

---

The problem A2A agents face

When two autonomous agents coordinate a financial action on Solana — one scanning signals, another executing a swap, a third approving the risk — each one needs to trust what the others saw. Right now there is no standard for that trust.

An agent can say "I checked the price before acting." But:

• Which oracle? At which slot?
• Was the data tampered with between source and decision?
• If something goes wrong, what exactly did the agent see?

Without answers to these questions, A2A coordination is a black box. Audits fail. Compliance is impossible. Debugging is guesswork.

What MIND adds

MIND is not a generic data API. It connects agents directly to native Solana data sources — Pyth oracle (on-chain price feeds) and Covalent/GoldRush (SPL token indexing) — and wraps every response in a mindprint: a SHA-256 proof of (inputs + response + Solana slot).

The slot anchor is the key. It ties the data to a specific moment in Solana's verifiable history. Any party can reconstruct and verify the proof independently.

Tool	Source	What it proves
mind_signals	Pyth oracle (on-chain)	Price + momentum at a specific slot
mind_market_intelligence	Covalent/GoldRush	Portfolio state at query time
mind_risk_scoring	MIND engine	Risk decision (PASS/WARN/BLOCK) with full input hash
mind_verify_proof	MIND proofs API	That a prior agent decision used unmodified data

This is the trust layer A2A protocols need to operate on Solana without a human in the loop for every action.

Install

git clone https://github.com/DGuedz/mind-intelligence-skill
cd mind-intelligence-skill
./install.sh

No API key required for x402 mode. The agent pays per call in USDC fractions automatically.

Two ways to use

Option 1 — x402 micropayments (no registration)

curl https://api.mindprotocol.xyz/v1/cards \
  -H "Content-Type: application/json" \
  -d '{"card": "mind_risk_scoring", "wallet": "7xKXtg2CW87d97TXJSDpbD5jBkheTqA83TZRuJosgAsU"}'

The agent receives a 402 Payment Required response with USDC payment instructions. Payment is automatic for any agent with a funded wallet. No signup, no API key, no dashboard.

Option 2 — API key (MCP / Claude Desktop)

export MIND_API_KEY=your_key_here
./install.sh

Get an API key: mindprotocol.xyz

Intelligence cards

Card	What it returns	Price
mind_signals	Pyth price + momentum at current Solana slot	0.010 USDC
mind_counterparty_check	Safe/Caution/Block verdict for a recipient wallet	0.015 USDC
mind_risk_scoring	Concentration risk score (0–100) + PASS/WARN/BLOCK	0.0154 USDC
mind_market_intelligence	Full SPL token balances + portfolio composition	0.020 USDC
mind_wallet_profile	Behavioral archetype + activity fingerprint	0.025 USDC
mind_token_intelligence	Holder concentration + insider flags + rug indicators	0.035 USDC
mind_portfolio_audit	Full portfolio risk report + DeFi exposure + token checks	0.10 USDC
mind_verify_proof	Verify any mindprint hash	free

Skill structure

skill/
  SKILL.md               ← entry point (routes both layers)
  signals.md             ← Pyth price signals
  portfolio.md           ← wallet/portfolio analysis
  counterparty.md        ← pre-tx counterparty check
  risk.md                ← risk scoring + PASS/BLOCK gate
  wallet-profile.md      ← behavioral fingerprint
  token-intelligence.md  ← token safety analysis
  portfolio-audit.md     ← full portfolio audit
  proofs.md              ← mindprint verification
  mcp.md                 ← MCP server setup
  x402/
    SKILL.md             ← x402 payment layer entry point
    docs/
      x402-overview.md   ← what x402 is and when to use it
      payment-flow.md    ← step-by-step lifecycle
      policy-gates.md    ← budget, allowlist, risk gates
      human-approval.md  ← HITL gate design
      proof-bundle.md    ← SHA-256 receipt hash chain
      rpc-reliability.md ← fallback and retry
agents/
  mind-trader-agent.md   ← reference A2A agent config
  x402-payment-agent.md  ← x402 payment agent config
commands/
  mind-check.md          ← /mind-check pre-flight command
examples/
  dry-run.md             ← x402 dry run (no real tx)

Usage in an A2A flow

An execution agent requests a risk check before swapping:

Check risk on wallet 7xKXtg2CW87d97TXJSDpbD5jBkheTqA83TZRuJosgAsU before I execute.

MIND calls mind_risk_scoring, returns PASS/BLOCK, and attaches a mindprint. The execution agent records the hash alongside the transaction. Later, any auditor can verify:

GET https://api.mindprotocol.xyz/v1/proof/{hash}

Returns the exact inputs and response the agent used — tied to the Solana slot at decision time. Not reconstructed. Not approximated. The original proof.

Why this is different from other skills in the kit

Skills like sendaifun/skills and jup-ag/agent-skills execute DeFi actions. MIND is the pre-flight layer that makes those actions auditable. It answers the question every A2A protocol eventually hits: how do you prove the agent acted on real, unmodified data?

No other skill in the kit produces tamper-evident audit trails per tool call — and no other skill accepts autonomous micropayments without requiring a human to register or manage an API key.

MIND Builder Program

Build a skill using MIND intelligence cards and earn 20% of every call it generates — paid weekly to your Solana wallet.

Open to all Superteam Brasil members and Solana builders globally.

See BUILDERS.md for details and registration →

License

MIT