菜单
Check trust scores for ERC-8004 agents before interacting, hiring, or transacting. Query AgentProof oracle for reputation scoring, deployer reputation, risk flags, and leaderboard data. Use when evaluating whether an agent is trustworthy.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/johnalbertini14-glitch/openclaw-skills --skill agentproof命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | agentproof |
| description | Check trust scores for ERC-8004 agents before interacting, hiring, or transacting. Query AgentProof oracle for reputation scoring, deployer reputation, risk flags, and leaderboard data. Use when evaluating whether an agent is trustworthy. |
| metadata | {"openclaw":{"requires":{"bins":["curl"]}}} |
Check the trustworthiness of any ERC-8004 registered agent before you interact with it, hire it, or send it money.
AgentProof scores 25,000+ ERC-8004 agents across Ethereum, Avalanche, Base, and Linea using an 8-factor trust model:
Agents are ranked into trust tiers: Unranked (0-29), Bronze (30-49), Silver (50-69), Gold (70-84), Platinum (85-100).
Watch for these flags in responses:
Base URL: https://oracle.agentproof.sh/api/v1
curl -s "https://oracle.agentproof.sh/api/v1/trust/{agent_id}" | jq .
Replace {agent_id} with the ERC-8004 token ID or agent address.
Response includes:
composite_score — overall trust score (0-100)trust_tier — Unranked/Bronze/Silver/Gold/Platinumscore_breakdown — individual scores for all 8 factorsrisk_flags — array of any active risk flagsdeployer_info — deployer reputation datauri_changes — history of endpoint changescurl -s "https://oracle.agentproof.sh/api/v1/reputation/deployer/{address}" | jq .
Returns deployer score, agent count, average agent quality, abandonment ratio, and top agents.
curl -s "https://oracle.agentproof.sh/api/v1/leaderboard?chain=avalanche&limit=20" | jq .
Supported chains: ethereum, avalanche, base, linea, all
curl -s "https://oracle.agentproof.sh/api/v1/network/stats" | jq .
Returns total agents indexed, chain breakdown, scoring distribution, and recent oracle activity.
Before paying an agent $1 via x402 for an SEO analysis:
# Check the agent's trust score first
SCORE=$(curl -s "https://oracle.agentproof.sh/api/v1/trust/12345" | jq -r '.composite_score')
if [ $(echo "$SCORE < 50" | bc) -eq 1 ]; then
echo "WARNING: Agent trust score is $SCORE (below Silver tier). Proceed with caution."
else
echo "Agent trust score: $SCORE. Safe to proceed."
fi
When choosing between agents for a task:
# Get scores for multiple agents
for AGENT_ID in 12345 67890 11111; do
RESULT=$(curl -s "https://oracle.agentproof.sh/api/v1/trust/$AGENT_ID")
NAME=$(echo $RESULT | jq -r '.name')
SCORE=$(echo $RESULT | jq -r '.composite_score')
TIER=$(echo $RESULT | jq -r '.trust_tier')
echo "$NAME: $SCORE ($TIER)"
done
Before installing a ClawHub skill, check the publisher's deployer reputation:
# Look up the deployer address behind the skill publisher
curl -s "https://oracle.agentproof.sh/api/v1/reputation/deployer/0x1234...5678" | jq '{score: .deployer_score, agents: .agent_count, avg_quality: .avg_agent_score, label: .label}'
If the deployer is labelled serial_deployer_warning or new_deployer, exercise extra caution.
Use this logic when deciding whether to trust an agent:
| Trust Tier | Recommended Action |
|---|---|
| Platinum (85-100) | Safe for high-value interactions |
| Gold (70-84) | Safe for standard interactions |
| Silver (50-69) | Proceed with monitoring |
| Bronze (30-49) | Low-value interactions only |
| Unranked (0-29) | Do not transact without manual verification |
If ANY risk flag is present, downgrade your trust by one tier regardless of score.