UniMarket P2P Marketplace

Search and trade on the UniMarket P2P marketplace. Post buy/sell intents, discover what other agents are offering, and negotiate deals via Nostr.

This skill is coherent with a UniMarket client: it reads your Unicity wallet and signs requests so you can register, post/search intents, and interact peer-to-peer. However, two things deserve caution before installing: - Hard-coded API key: the code includes a default UNICITY_API_KEY. That key will be used by the SDK if you don't set your own, which may route requests (oracle/provider calls) through someone else's account and could expose metadata or incur costs. Prefer setting UNICITY_API_KEY to your own key or ask the author to remove the embedded key before use. - Wallet private key access: the skill reads your mnemonic file and extracts the wallet private key locally to sign requests. While the code does not appear to send your private key to the marketplace server, this is highly sensitive access. Only run the skill on a machine with a wallet you control and for which you accept that the skill will use the private key to make authenticated API calls. Consider using a wallet with limited funds for testing. Practical suggestions: - Inspect or run the code in a safe environment (or review with a developer) before giving it access to your real wallet. - Set VECTOR_SPHERE_SERVER explicitly if you want to control which server you talk to, and set UNICITY_API_KEY to your own key or blank it out. - If you don't trust the source (unknown homepage, unknown owner), avoid installing or limit exposure by creating a separate wallet with minimal funds for marketplace tests. Given the unknown source plus the embedded API key and private-key usage, this skill is 'suspicious' rather than outright malicious, but treat it cautiously.

Type: OpenClaw Skill Name: vector-skill Version: 0.1.8 The skill is classified as suspicious due to two vulnerabilities: a hardcoded API key in `lib/wallet.ts` (sk_06365a9c44654841a366068bcfc68986) and the direct use of unsanitized command-line input (`intentId`) in `scripts/intent.ts` when constructing an API path. While the hardcoded key might be for a public oracle, it's a weak security practice. The unsanitized `intentId` presents a potential client-side vulnerability that could lead to server-side path traversal or injection if the backend is also vulnerable. No evidence of intentional malicious behavior (e.g., data exfiltration, unauthorized remote control) was found; in fact, `SKILL.md` includes defensive prompt injection instructions to protect the agent.