Add ENS agent identity skill

[estmcmxci]

Summary

• Adds ens-agent-identity skill: ENS-based identity layer for Bankr agents
• Human-readable names via NameStone (gasless offchain subnames, CCIP-Read)
• Structured metadata via agent:* text records (type, capabilities, chains, a2a endpoint, etc.)
• ENSIP-25 verification linking ENS names to ERC-8004 on-chain identity on Base
• Three scripts: set-agent-records.sh, resolve-agent.sh, verify-agent-registration.sh

Why

Bankr agents are currently API keys + wallet addresses — no names, no discoverability, no portable reputation. This skill gives agents human-readable ENS names with structured metadata, and bridges them to ERC-8004 on-chain identity via ENSIP-25 verification.

Tested end-to-end on mainnet

• Registered alpha-go.bankrtest.eth via NameStone (CCIP-Read resolution confirmed)
• Registered agent #19327 in ERC-8004 Identity Registry on Base (tx)
• ENSIP-25 full verification: ENS name linked to ERC-8004 identity confirmed

Context

Part of a broader ENS x Bankr integration proposal covering three phases:

1. ENS subnames via NameStone (this PR)
2. Agent metadata via text records (this PR)
3. ERC-8004 identity & reputation (roadmap)

Infrastructure: NameStone for gasless subname management (built by slobo.eth, ENS Ecosystem steward). Namespace uses mainnet ENS + CCIP-Read (chain-neutral, not pinned to Base).

Test plan

• Review SKILL.md prompt guide
• Review agent:* text record schema
• Review ENSIP-25 verification spec and ERC-7930 encoding
• Run set-agent-records.sh with a NameStone API key
• Run resolve-agent.sh against alpha-go.bankrtest.eth
• Run verify-agent-registration.sh against agent #19327 on Base

🤖 Generated with Claude Code

[estmcmxci]

Code review

Found 1 issue:

1. ABI decoding offset bug in verify-agent-registration.sh permanently breaks ERC-8004 verification. hex.slice(130, 194) reads the first 32 bytes of string data as the length field, instead of the actual length at hex.slice(66, 130). For a tokenURI(uint256) return, the ABI layout after the 0x prefix is: offset (chars 2-65), length (chars 66-129), data (chars 130+). The current code interprets URI bytes as a huge integer for length, then reads past the end of the buffer. Result: REGISTRY_RESULT is always empty, ERC8004_VERIFIED is always false.

https://github.com/BankrBot/openclaw-skills/blob/cd1cadb01db0884190243f1d92c5016921a4c88f/ens-agent-identity/scripts/verify-agent-registration.sh#L155-L163

Fix:

const length = parseInt(hex.slice(66, 130), 16);
if (length > 0) {
  const str = Buffer.from(hex.slice(130, 130 + length * 2), 'hex').toString('utf8');

🤖 Generated with Claude Code

_{- If this code review was useful, please react with 👍. Otherwise, react with 👎.}