across-protocol-ai-agent-skill

Warn

Audited by Socket on Sep 25, 2026

2 alerts found:

Securityx2
SecurityMEDIUM
swap/SKILL.md

BENIGN but high-impact. The skill is coherent with Across swap integration and uses official API domains, with no suspicious installer or proxy behavior. Main risk is that it enables financially consequential onchain actions, especially approvals, swaps, and embedded destination-chain calls; this warrants high operational caution even though there is little evidence of malicious intent.

Confidence: 86%Severity: 72%
SecurityMEDIUM
embedded-crosschain-actions/SKILL.md

SUSPICIOUS: The skill is legitimate Across documentation with official same-org data flows and no supply-chain red flags, but it grants an AI agent broad ability to trigger arbitrary crosschain and destination-chain financial actions. Risk comes from high-impact autonomous blockchain execution, not from malware or credential theft behavior.

Confidence: 90%Severity: 74%
Audit Metadata
Analyzed At
Sep 25, 2026, 08:58 PM
Package URL
pkg:socket/skills-sh/across-protocol%2Fskills%2Facross-protocol-ai-agent-skill%2F@55c4c78d6b66f0325fd101b3ca63884a82a53dea681644b45fefc509ddea2818