AI Bounty Claim Skill

中文

Quick Start

Send the prompt below to your AI. Replace the email placeholder with your own email first, and do not use an exchange or custodial address.

Prompt for Your AI

1. Please read https://github.com/aelf-hzz780/AI-bounty-skill/blob/main/SKILL.md, understand this skill, and install it locally.
2. Then help me claim by following the skill instructions, and use the AA/CA route.
3. Use email: <replace-with-your-email> to recover/login.

Fast Paths by Host

Use a host-specific entry only when it maps to a real native install or discovery path. In this repository today, Gemini CLI is the only true one-command install, OpenClaw and OpenCode use native workspace skill discovery, and Claude Code plus Codex use bootstrap install docs.

Gemini CLI

This repository now ships gemini-extension.json and GEMINI.md, so Gemini CLI can install it as an extension:

gemini extensions install https://github.com/aelf-hzz780/AI-bounty-skill

OpenClaw

This repository now ships a workspace skill at skills/ai-bounty-claim/SKILL.md. Open this repo as your OpenClaw workspace and start a new session.

OpenCode

This repository now ships a project skill at .opencode/skills/ai-bounty-claim/SKILL.md. Open the repo in OpenCode and start a new session.

Claude Code

Tell Claude Code:

Fetch and follow instructions from https://raw.githubusercontent.com/aelf-hzz780/AI-bounty-skill/main/.claude/INSTALL.md

Codex

Tell Codex:

Fetch and follow instructions from https://raw.githubusercontent.com/aelf-hzz780/AI-bounty-skill/main/.codex/INSTALL.md

Cursor

No dedicated fast path yet. Use the universal prompt above for now. Add a Cursor-specific entry only after this repo has .cursor/rules or a real MCP install path.

This public repository contains a single skill for claiming the AI bounty on the aelf tDVV mainnet sidechain through RewardClaimContract.

For AA/CA, the standard wallet claim path is manager signer -> CA.ManagerForwardCall -> reward.ClaimByPortkeyToCa(Hash ca_hash).

Compatibility

• Skill version in this repo: 3.0.0
• AA/CA dependency: @portkey/ca-agent-skills >= 2.3.0
  • required interfaces: CLI manager-sync-status, MCP portkey_manager_sync_status
• Release status: @portkey/ca-agent-skills 2.3.0 is already published, so this repo no longer treats CA release timing as a blocker for AI-bounty 3.0.0

Change Log

3.0.0

• Breaking change: this public skill no longer supports the legacy EOA claim route.
• Removed: EOA onboarding guidance, EOA reward copy, EOA gas guidance, and the dedicated EOA flow/example files.
• Current public claim path is AA/CA only: ManagerForwardCall -> ClaimByPortkeyToCa.
• Older reward contract source snapshots may still show legacy EOA methods, but this repository no longer exposes them as supported user-facing paths.

Repository Design

The repository uses a single entry skill plus focused branch references so weaker agents can follow one route at a time.

• SKILL.md: onboarding, routing rules, hard stops, and shared defaults
• references/flows/: branch-specific step-by-step instructions
• references/examples/: branch-specific examples for weaker models

Supported Claim Paths

This skill supports only the current public claim path:

• AA/CA: ManagerForwardCall(...) -> ClaimByPortkeyToCa(Hash ca_hash)

Routing Branches

• Account choice and onboarding
• Portkey AA/CA claim
• Diagnostics and stop

Shared Rules

• The canonical skill version is the version field in SKILL.md. Ask external users or agents to report that value first when behavior looks inconsistent.
• tDVV is documented here as the current AI bounty mainnet sidechain environment.
• Current campaign default reward amount is documented as 2 AIBOUNTY for AA/CA.
• AA/CA claims should use the standard wallet path manager signer -> CA.ManagerForwardCall -> reward.ClaimByPortkeyToCa(Hash ca_hash).
• Before any AA/CA write, the agent must follow the fixed read-only gate resolve local context and caHash -> manager-sync-status / portkey_manager_sync_status -> GetHolderInfo(caHash) -> explicit confirmation.
• The AA/CA write path must stop unless isManagerSynced=true and GetHolderInfo(caHash) resolves on tDVV.
• ClaimByPortkeyToCa(Hash ca_hash) is permissionless at the reward method layer, but the reward still goes to the resolved caHash -> caAddress, not to the manager signer.
• For AA/CA SDK or helper calls that require raw addresses, use raw CA and reward addresses instead of wrapped ELF_..._tDVV addresses.
• If only an email is known for AA/CA, the agent should resolve the target caHash first, and recover a usable manager signer when needed for the standard wallet path.
• Current environment gas rules are documented as daily subsidy behavior around 1 ELF; for AA/CA, the agent should usually allow one confirmed attempt on the standard wallet path even when visible ELF is low or zero, but only after the AA/CA sync and holder gate passes.
• Holder is not found and CA holder not found must be mapped to holder, caHash, or target-chain context not being ready. The agent must not infer a fee problem or suggest sending 0.01 ELF.
• Sender is not a manager of the CA holder. must be mapped to manager sync or local manager-context mismatch on tDVV.
• Transaction fee not enough. must be treated as a send-stage fee problem only, not as a holder lookup problem.
• NOTEXISTED means pending lookup only and must not be treated as final failure or a fee blocker.
• Validate RPC reachability with chainStatus, not by requesting the RPC root URL.
• Do not use to conclude that the reward contract lacks write methods.

Usage

Start from SKILL.md, choose exactly one branch, then read the matching flow and example documents before replying.