Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Submit to AgentBeat
v1.9.0Complete submission flow for autonomous AI agents into the on-chain agent economy. Guides agents through creating an EVM wallet, requesting gas from their ow...
⭐ 0· 1.3k·1 current·1 all-time
by@togodn2
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
SKILL.md content (wallet creation, gas requests, ERC‑8004 registration, x402 integration, AgentBeat submission) aligns with the skill name and description. However, the published registry metadata claims no required binaries, no env vars, and no config paths, while the SKILL.md header lists required tools (node, npm, curl, optional jq), an optional EVM_PRIVATE_KEY env var, and a credentials_path (~/.config/agentbeat/credentials.json). This mismatch is a packaging/inventory inconsistency that should be clarified.
Instruction Scope
Runtime instructions are explicit and largely constrained to the stated purpose: generating/using an EVM key, verifying endpoints, calling RPCs, interacting with x402 facilitators and AgentBeat API, and writing a credentials.json to the user's home. The skill includes mandatory owner gates before sensitive actions. Still, examples show generating and printing private keys to stdout and describe persisting credentials to ~/.config/agentbeat/credentials.json — both are sensitive operations. The instructions require network access to multiple external endpoints (mainnet RPCs, facilitator.world.fun, api.agentbeat.fun), which is expected but important to surface to the owner.
Install Mechanism
No install spec or code files are included; this is an instruction-only skill. That limits the surface (nothing is downloaded or written by an installer). Example commands reference npm/pip/openssl, but no automated installer is provided.
Credentials
Only one optional credential (EVM_PRIVATE_KEY) is mentioned in SKILL.md which is appropriate for on‑chain signing. The skill will write to ~/.config/agentbeat/credentials.json and store non-secret metadata plus sensitive items (voucher, possibly private key if owner approves). The SKILL.md enforces owner approval before storing plaintext keys, which is good practice, but persisting any private key or voucher to disk is inherently high-risk and justified only by the skill's purpose. Registry metadata did not advertise the credentials path or required tools, creating a discrepancy.
Persistence & Privilege
The skill requests persistent storage of a credentials file in the user's home directory to save addresses, agentId, voucher, and decision records. It does not request always:true or try to modify other skills. Persisting the voucher and non-secret metadata is reasonable for the workflow; storing a plaintext private key is allowed only after explicit owner approval per the documented gates.
What to consider before installing
This skill appears to implement the advertised AgentBeat submission flow, but check these things before installing or running it:
- Packaging mismatch: SKILL.md lists required tools (node, npm, curl, optional jq), an optional EVM_PRIVATE_KEY env var, and a credentials path (~/.config/agentbeat/credentials.json), but the registry metadata claims none of these — confirm which is accurate before running.
- Private key risk: Examples show generating and printing private keys; do NOT store private keys in plaintext unless the owner explicitly approves and understands the risk. Prefer an external signer, OS keychain, or encrypted vault as the SKILL.md recommends.
- Credentials file: The skill writes ~/.config/agentbeat/credentials.json and may store the submission voucher there. Ensure the file is protected (chmod 600) and do not commit it to version control.
- External endpoints: The flow communicates with mainnet RPC endpoints, AgentBeat API, and x402 facilitator services (e.g., facilitator.world.fun). Verify you trust those endpoints and their privacy/security properties before allowing the agent to interact with them.
- Voucher sensitivity: The submission voucher is required to claim rewards and is sensitive; do not post it publicly without explicit owner approval.
- Operational checks: The skill enforces owner confirmation gates (agent legitimacy, endpoint declaration, reward address, ownership proof). Use those gates; do not bypass them.
If you plan to use this skill, ask the skill author/maintainer to fix the metadata inconsistencies (declare required tools/env/config path in registry metadata) and review/confirm the exact runtime operations you will allow (especially any decision to persist private keys).Like a lobster shell, security has layers — review code before you run it.
latestvk97c36m4r420nndxmft5md7q3x82nqcs
License
MIT-0
Free to use, modify, and redistribute. No attribution required.