opensea

SUSPICIOUS: The skill is broadly coherent with its stated OpenSea tool-registry purpose and uses mostly official endpoints and npm distribution, so it does not look malicious. Risk is medium because it can execute onchain transactions, make paid calls, and optionally forward wallet credentials/signing to third-party providers, which is a high-trust capability for an AI agent skill.

SUSPICIOUS: the skill is purpose-aligned and mainly uses official OpenSea infrastructure, so it does not look malicious. However, it enables autonomous high-impact financial transactions and depends on external wallet-signing providers, making the overall security risk high despite low evidence of malware.

BENIGN for stated purpose but high-impact. The skill is internally consistent and uses official OpenSea channels, with no clear exfiltration or deceptive install path. Main risk comes from enabling autonomous financial transactions and from credential/signing trust delegated to wallet providers during swap execution.

No direct malicious payload is evident in this YAML fragment (no obfuscation, no obvious exfiltration logic, no shell backdoors). The predominant security concern is supply-chain and CI/CD integrity: the workflow installs and executes a third-party npm CLI at runtime with access to a sensitive token, then publishes the checked-out repository directory to an external registry. This creates a high-impact target—if the CLI, its dependencies, or the repository content/workflow inputs are compromised, unintended or malicious artifacts could be published and the token could potentially be misused. Recommended hardening (not shown here) would include pinning actions by commit SHA, verifying the clawhub CLI provenance/integrity (e.g., lockfiles/signatures), and ensuring only trusted repository contents/tags can trigger publishing.