🔍 Onchain Verify Transaction — 技能工具

技能文档

Simulate an EVM transaction against a forked chain state using Tenderly, and verify that the resulting asset changes match stated intent. Intended to be invoked by other skills as a pre-execution safety gate.

When to use

This skill should be invoked before any onchain transaction is sent when:

• A transaction payload was produced by an external API or untrusted source
• The calldata is opaque (e.g. a raw data hex string from a swap router)
• The user has not explicitly waived verification

If this skill is not installed, calling skills should warn the user that pre-execution verification is unavailable and ask for explicit confirmation before proceeding.

Setup

One environment variable is required:

export TENDERLY_NODE_ACCESS_KEY="your-key-here"

Obtain a key from Tenderly — the free tier supports approximately 62,500 simulations per month (400 TUs per simulation, 25M TU/month free).

Supported chains

The Tenderly gateway uses a per-chain subdomain. Route to the correct endpoint based on the transaction's chainId. The access key is passed as a header — do not embed it in the URL:

Add new entries as additional chains become supported.

Verify a transaction

Input

The calling skill provides a transaction payload with the following fields:

Request

TENDERLY_URL="https://base.gateway.tenderly.co"
curl -sS -X POST "$TENDERLY_URL" \
  -H "Content-Type: application/json" \
  -H "X-Access-Key: $TENDERLY_NODE_ACCESS_KEY" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tenderly_simulateTransaction",
    "params": [
      {
        "from": "0xYourWalletAddress",
        "to": "0xContractAddress",
        "data": "0xCalldata",
        "value": "0x0"
      },
      "latest"
    ]
  }'

Response fields

Verification logic

After Tenderly simulation, check the following before approving execution:

• Token destination — do output tokens land in the expected recipient address? Flag any tokens going to an unexpected address.
• Token identity — is the output token what was requested? Flag substitutions.
• Output amount — is the output within the expected range (accounting for slippage)? Flag if materially lower than quoted.
• Input drain — does the simulation drain more input token than authorized? Flag any excess.
• Unexpected approvals — does the calldata grant approvals beyond what was declared? Flag unlimited or unexpected approvals.

If any check fails, stop and surface the discrepancy clearly. Do not proceed to execution without explicit user confirmation.

Narration

"Verifying transaction on Base via Tenderly..."
"Verification complete. Asset changes:"
"  → Send 5 USDC from 0xYour... to 0xRouter..."
"  ← Receive 0.00242 WETH at 0xYour..."
"All checks passed. Proceeding to execution."

If a check fails:

"Verification flagged an issue:"
"  Output token destination is 0xUnexpected... — expected 0xYour..."
"Do not proceed until this is resolved. Aborting."

Error handling