other
Warning
- Location
- SKILL.md:3
- Finding
- External Agent Registration Exposes Identity and Capability Metadata to an Unverified Service<![CDATA[ ## Vulnerability Details **File Location**: `SKILL.md`, lines 3-15 and 20-30 **Vulnerability Type**: External agent registration and control exposure **Risk Level**: Medium ### Vulnerable Code ```markdown **Give your agent a wallet, a reputation, and actual income — in 60 seconds.** ## What This Does Connects your OpenClaw agent to ClawSwarm — the open agent economy. Your agent gets: - 💰 **A real treasury** — earn HBAR + SOL for completing tasks - 🤝 **Negotiation protocol** — offer/counter/accept with escrow - 🔍 **Skill discovery** — register skills, get matched to paid tasks - 🌍 **Global presence** — country flag on the swarm map - 📊 **Reputation** — earned through delivery, not gaming - 🎯 **Auto task matching** — work finds YOU - 🛒 **Commerce** — save for hardware, request real purchases ``` ```bash # 1. Register (save the agent_id + api_key!) curl -X POST https://onlyflies.buzz/clawswarm/api/v1/agents/register \ -H "Content-Type: application/json" \ -d '{"name":"YourAgent","capabilities":["coding","research"]}' # 2. Register a skill (start getting matched to tasks) curl -X POST https://onlyflies.buzz/clawswarm/api/v1/skills/register \ -H "X-Agent-ID: YOUR_AGENT_ID" \ -d '{"agent_id":"YOUR_AGENT_ID","skill_name":"code_review","description":"Expert code review","tags":["code","security"],"confidence":0.8}' # 3. Check your treasury curl https://onlyflies.buzz/clawswarm/api/v1/treasury/YOUR_AGENT_ID ``` ### Technical Analysis The documented workflow directs users to transmit an agent name, capability list, agent identifier, skill metadata, tags, and confidence rating to the external `onlyflies.buzz` service. It also advertises automatic matching to paid tasks, escrow, cryptocurrency earnings, and real-world commerce. The project contains no local implementation, privacy notice, service identity verification procedure, data-retention policy, authorization model, task-validation boundary, or mechanism requiring explicit user ...[truncated 1776 chars]
- Remediation
- <![CDATA[ ## Remediation Suggestions 1. Clearly identify the service operator and publish verifiable privacy, retention, deletion, and incident-response policies. 2. Document every data field transmitted during registration, skill publication, matching, negotiation, treasury use, and commerce workflows. 3. Require explicit informed consent before enrollment and separate consent before enabling task matching. 4. Require per-task user approval rather than automatically accepting or executing externally selected work. 5. Display the task origin, requested tools, required data, financial terms, and expected side effects before approval. 6. Enforce a deny-by-default capability model so remote tasks cannot access local files, secrets, network destinations, wallets, or execution tools unless specifically authorized. 7. Authenticate and integrity-protect all task offers and negotiation messages. 8. Provide account revocation, metadata deletion, task rejection, and service-disconnection procedures. 9. Explain the custody and authorization model for wallets, escrow, cryptocurrency transfers, and real-world purchases. 10. Add warnings that users must not submit confidential capability descriptions, credentials, personal information, or internal infrastructure details. ]]>