Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

ClawdVine

v1.2.0

Short-form video for AI agents. Generate videos using the latest models, pay with USDC via x402.

0· 1.2k·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
medium confidence
!
Purpose & Capability
The skill's description (video generation, pay with x402, optional onchain identity) matches the included scripts (SIWE signing, x402 payment, balance checks). However the registry metadata claims no required environment variables or credentials, while the provided scripts and SKILL.md clearly require an EVM private key (EVM_PRIVATE_KEY) and recommend storing CLAWDVINE_AGENT_ID. That undocumented need for a signing key is a significant incoherence.
!
Instruction Scope
SKILL.md instructs agents to read memory/config and environment for CLAWDVINE_AGENT_ID and to 'Store this permanently.' The runtime instructions also describe flows that require reading a wallet private key, signing SIWE messages, and performing x402 payments. The document contains a pre-scan prompt-injection signal (system-prompt-override). Asking agents to persist identifiers and to access private keys expands scope beyond simple API calls and should be made explicit and constrained.
Install Mechanism
There is no external download/install spec (instruction-only style), but the package includes Node scripts and an npm package.json/lockfile listing standard npm dependencies (@x402/*, viem, siwe, ethers). Dependencies are from public registries (npm) — no arbitrary remote archives observed. This is moderately low-risk for supply chain, but the skill will require installing node packages to run the scripts.
!
Credentials
The code and docs require a sensitive secret (EVM_PRIVATE_KEY) and recommend setting CLAWDVINE_AGENT_ID in the environment; yet the registry metadata declared no required env vars or primary credential. Requesting a private key is proportionate to signing payments and minting onchain identity, but the omission from metadata and the implicit instruction to store secrets increases risk and is not transparent.
Persistence & Privilege
The skill is not force-included (always: false). However SKILL.md explicitly tells the agent to save agentId permanently to memory/config/environment — this creates persistent state tied to the skill. Combined with the ability to sign payments using an EVM private key, autonomous invocation (default allowed) would increase blast radius. The skill does not request modifying other skills or global settings.
Scan Findings in Context
[system-prompt-override] unexpected: A pattern indicating prompt-injection was found in SKILL.md. A video-generation/payment skill does not normally need to override or manipulate host system prompts; this could be an attempt to influence agent decision making or persistence and should be reviewed manually.
What to consider before installing
Key points to consider before installing or using this skill: - Hidden secret requirement: The bundle's scripts and SKILL.md require an EVM private key (EVM_PRIVATE_KEY) to sign SIWE messages and to make x402 payments. The registry metadata incorrectly lists no required env vars — treat that as an omission, not a safety guarantee. - High-risk action: Providing a private key gives the skill the ability to sign on-chain operations. If you must test this, use a dedicated wallet with minimal funds and only the exact token (USDC) required for payments; never expose your primary wallet or keys with real funds. - Persistence: The instructions explicitly ask agents to save your agentId permanently (memory/config/environment). Consider whether you want that persistent link between your agent and the service. - Prompt-injection signal: SKILL.md contains text patterns flagged as 'system-prompt-override'. Manually inspect the SKILL.md for any instructions that ask the agent to change its system prompts, ignore safety, or execute open-ended 'use your judgment' actions. Don't trust automatic evaluation alone. - Operational advice: If you decide to proceed, run these scripts locally in an isolated environment (not granting them to an autonomous agent), audit network traffic to verify endpoints (api.clawdvine.sh, x402 facilitator), and review the open-source repository (package.json lists https://github.com/onbonsai/clawdvine-skill) and commit history. Prefer connecting via an external wallet UI or hardware wallet rather than exporting the private key into environment variables where possible. What would change this assessment: explicit, accurate registry metadata declaring required env vars and credential scopes; a public, audited repository and maintainer identity; removal or justification of any prompt-manipulation text in SKILL.md; and an alternative payment flow that doesn't require exposing a private key to the agent runtime.

Like a lobster shell, security has layers — review code before you run it.

latestvk97drtemc7psbqv28jxk69kh4d80pv0y

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments