Musashi

Conviction-weighted token intelligence. Analyze any token through 7 elimination gates, cross-domain pattern detection, and adversarial debate. Triggers on "a...

Key things to check before installing and running: - Review source before building. The repo includes full Go code that will be compiled into a binary; compile only after you (or someone you trust) inspects it. - Do not set OG_CHAIN_PRIVATE_KEY in a global/shared config. If you want to enable publishing, use a dedicated, minimal-balance wallet (as SKILL.md recommends) and inject the private key only when you intentionally perform a publish action. - Confirm whether you actually need to set CONVICTION_LOG_ADDRESS and MUSASHI_INFT_ADDRESS for your use case. The skill declares them as required, but read-only analysis should not require them — if you must set them, verify the addresses are correct and intended. - Understand that the binary exposes explicit commands (strike, mint-agent, update-agent, store) that will sign/send transactions if a private key is present; the SKILL.md policy to ask for confirmation is a human/process control, not a programmatic guard in the binary. - Because the tool makes many network/API calls (DexScreener, CoinGecko, DefiLlama, GoPlus, Farcaster optional), expect outbound network traffic — if you have network egress policies, review them. If you want lower risk: run the built binary in a constrained environment (no OG_CHAIN_PRIVATE_KEY, network egress limited to required APIs) and exercise analysis-mode gates only. If you plan to enable publish functionality, perform an independent code review and use a dedicated signing wallet with minimal funds.

Type: OpenClaw Skill Name: musashi Version: 2.0.0 The 'musashi' skill bundle implements a sophisticated crypto token analysis engine with a 7-gate pipeline and on-chain reporting. It possesses high-risk capabilities, specifically the handling of a private key (OG_CHAIN_PRIVATE_KEY) to sign transactions on the 0G Chain and the execution of shell commands to interact with a local Go binary and the '0g-storage-client' CLI. While the SKILL.md instructions include strong safety constraints—such as mandatory user confirmation before analysis and publishing, and advice to use a dedicated wallet—the presence of these powerful capabilities without a sandbox environment is inherently risky. The code logic in 'conviction.go' and 'inft.go' for transaction signing and 'og_storage.go' for file uploads appears aligned with the stated purpose, but the bundle is classified as suspicious due to the high-risk nature of its core functions.