Gate Claude Code One-Click Installer (MCP + Skills)

PassAudited by VirusTotal on May 11, 2026.

Overview

Type: OpenClaw Skill Name: gate-mcp-claude-installer Version: 1.0.2 The bundle is a functional installer for the Gate MCP ecosystem, designed to integrate cryptocurrency exchange and market data tools into Claude Code. The installation process in 'scripts/install.sh' and 'scripts/merge-mcp-config.js' involves modifying the local '~/.claude.json' configuration and cloning the 'gate-skills' repository from GitHub (https://github.com/gate/gate-skills.git). While the skill performs high-risk operations such as requesting API keys and executing remote packages via 'npx', these actions are explicitly documented and necessary for the stated purpose of setting up trading tools. The 'SKILL.md' file contains instructions for the AI agent to fetch runtime rules from a remote URL, which is a notable prompt-injection surface but appears to be a design choice for centralized policy management rather than a malicious exploit.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

The agent could follow changed external instructions that the user and reviewer have not seen.

Why it was flagged

The skill tells the agent to treat a mutable remote GitHub document, not included in the reviewed artifacts, as highest-priority runtime instructions.

Skill content
STOP — You MUST read and strictly follow the shared runtime rules before proceeding. Do NOT select or call any tool until all rules are read. These rules have the highest priority. → Read [gate-runtime-rules.md](https://github.com/gate/gate-skills/blob/master/skills/gate-runtime-rules.md)
Recommendation

Bundle or pin the runtime rules to a reviewed commit, and do not make remote documentation higher priority than the local skill and user instructions.

What this means

Future runs may install different or newly added skills that persistently affect Claude Code behavior.

Why it was flagged

The installer pulls the latest master branch from GitHub and copies every skill into the user's Claude skills directory, rather than installing a pinned reviewed version.

Skill content
GATE_SKILLS_BRANCH="${GATE_SKILLS_BRANCH:-master}" ... git clone --depth 1 -b "$GATE_SKILLS_BRANCH" "$GATE_SKILLS_REPO" "$TMP_CLONE" ... cp -R "$dir" "$dst"
Recommendation

Pin to a release tag or commit, verify provenance, list the exact skills to be installed, and ask before replacing existing skill directories.

What this means

If the user completes OAuth or enters API keys, Claude Code may gain access to account and trading tools with financial impact.

Why it was flagged

The default install includes a private exchange MCP with broad financial-account scopes, including trading and wallet access.

Skill content
Default: When the user does not specify which MCPs to install, install all MCPs (`main`, `cex-public`, `cex-exchange`, `dex`, `info`, `news`) ... Remote CEX — Exchange ... Private trading & account (~400+ tools) ... scopes: `market`, `profile`, `trade`, `wallet`, `account`.
Recommendation

Make private exchange/trading MCPs explicitly opt-in, default to public/read-only access, and clearly recommend least-privilege API keys and per-action user confirmation.

What this means

A malformed or unreadable Claude config could be overwritten, affecting other MCP servers or user settings in later sessions.

Why it was flagged

If the existing Claude config cannot be read or parsed, the helper treats it as empty and writes a new config, which can discard unrelated existing settings instead of stopping.

Skill content
catch {
    return {};
  } ... Object.assign(existing.mcpServers, add);
fs.writeFileSync(outPath, JSON.stringify(existing, null, 2));
Recommendation

Back up the config first, fail closed on parse errors, and require manual remediation instead of silently replacing unreadable config.