codexmcp-claude-codex-collaboration
Warn
Audited by Socket on May 17, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
SUSPICIOUS. The skill's general purpose is coherent, but its trust chain is not: an ara.so skill instructs the agent to execute a GitHub-hosted MCP server from a different publisher via `uvx`, then route project data and possibly credential-backed Codex activity through it. The capability belongs to the stated purpose, but the install provenance, auto-approval guidance, and optional high-autonomy modes make the overall skill medium-high risk rather than benign.
Confidence: 84%Severity: 72%
Audit Metadata