registry  /  amicus  /  2.1.0

amicus@2.1.0

Multi-model LLM Council + parallel AI window for Claude Code. Run structured council reviews across Gemini, GPT, DeepSeek and more — or fork a conversation to any model and fold the results back.

AI Security Review

scanned 3h ago · by lpm-firewall-ai

LPM blocks this version under the AI-agent control-surface policy. Install-time code mutates broad Claude agent control surfaces without an explicit user invocation. It plants Claude skills and registers a standing MCP server that later launches amicus via npx latest.

Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review; previous stored version diff introduced dangerous source
Trigger
npm install / npm install -g package lifecycle postinstall
Impact
Claude Code/Desktop gains package-supplied skills and an auto-registered MCP server capable of spawning Amicus/OpenCode agent sessions from the user/home context.
Mechanism
unconsented lifecycle mutation of Claude skills and MCP configs
Policy narrative
On installation, the package runs scripts/postinstall.js. That script copies package-authored skills into the user's global Claude skills directory and registers an MCP server named amicus in Claude Code and Claude Desktop/Cowork configs, using npx -y amicus@latest mcp. The content is product-aligned, but the delivery is an npm lifecycle hook mutating foreign/broad AI-agent control surfaces without a separate user-invoked setup step.
Rationale
Static inspection confirms the scanner's critical finding: npm postinstall writes Claude skills and MCP registrations into user-level Claude control surfaces. Under the provided install-control-surface policy, unconsented lifecycle delivery into a foreign/broad agent surface is blockable even when product-aligned.
Evidence
package.jsonscripts/postinstall.jsskills/sidecar/SKILL.mdskills/second-opinion/SKILL.mdsrc/mcp-server.jssrc/mcp-tools.js~/.claude/skills/sidecar/SKILL.md~/.claude/skills/second-opinion/SKILL.md~/.claude/skills/second-opinion/COUNCIL-DESIGN.md~/.claude/skills/second-opinion/MODEL-NOTES.md~/.claude.json~/Library/Application Support/Claude/claude_desktop_config.json%APPDATA%/Claude/claude_desktop_config.json~/.config/claude/claude_desktop_config.json

Decision evidence

public snapshot
AI called this Malicious at 96.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for policy block
  • package.json defines postinstall: node scripts/postinstall.js.
  • scripts/postinstall.js copies package skills into ~/.claude/skills/sidecar and ~/.claude/skills/second-opinion during npm install.
  • scripts/postinstall.js registers an amicus MCP server in Claude Code via claude mcp add-json or ~/.claude.json fallback.
  • scripts/postinstall.js writes Claude Desktop/Cowork claude_desktop_config.json with command npx -y amicus@latest mcp.
  • skills/sidecar/SKILL.md instructs Claude to run amicus/npx commands and use background launches; this is planted into Claude's global skill surface by lifecycle code.
Evidence against
  • Package purpose is openly agent/LLM sidecar and council tooling, and the planted skill/MCP content is product-aligned.
  • Postinstall has AMICUS_SKIP_POSTINSTALL=1 opt-out and tries to preserve existing amicus-shaped MCP env values.
  • Default Electron postinstall provisioning is cache-only; network prewarm requires AMICUS_PREFETCH_ELECTRON=1.
  • No credential exfiltration or destructive file deletion found in inspected source.
Behavioral surface
Source
ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNetwork
Supply chain
HighEntropyStringsUrlStrings
ManifestNo manifest risk signals triggered.
scanned 146 file(s), 908 KB of source, external domains: aistudio.google.com, api.anthropic.com, api.deepseek.com, api.openai.com, console.anthropic.com, generativelanguage.googleapis.com, nodejs.org, openrouter.ai, platform.deepseek.com, platform.openai.com, www.w3.org

Source & flagged code

7 flagged · loading source
package.jsonView file
scripts.postinstall = node scripts/postinstall.js
High
Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg
scripts.postinstall = node scripts/postinstall.js
Medium
Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg
bin/amicus.jsView file
10// Load API keys from all sources: process.env > amicus .env > auth.json L11: const { loadCredentials } = require('../src/utils/env-loader'); L12: loadCredentials();
Medium
Dynamic Require

Package source references dynamic require/import behavior.

bin/amicus.jsView on unpkg · L10
scripts/postinstall.jsView file
5Install-time AI-agent control hijack evidence: L5: * L6: * 1. Copies SKILL.md to ~/.claude/skills/sidecar/ L7: * 2. Registers MCP server in Claude Code (~/.claude.json) L8: * 3. Registers MCP server in Claude Desktop/Cowork config ... L37: function skillsRoot() { L38: return path.join(os.homedir(), '.claude', 'skills'); L39: } ... L70: L71: if (!existing.mcpServers) { existing.mcpServers = {}; } L72: L73: const prev = existing.mcpServers[name]; L74: const nextConfig = (prev && isAmicusMcpConfig(prev)) ? { ...prev, ...config } : config; Payload evidence from skills/sidecar/SKILL.md: L41: 1. **ALWAYS launch amicus CLI commands with the Bash tool's `run_in_background: true`.** Never run `amicus start/resume/continue` in the foreground. L42: 2. **The fold summary returns on stdout** when the user clicks Fold in the GUI or the headless agent finishes. Use TaskOutput to read it when the background task completes. L43: 3. **For long or multi-line briefings, write them to a temp file and pass `--prompt-file <path>`** (mutually exclusive with `--prompt`; avoids shell-quoting hazards and argument-si... ... L82: **Step 1: Get an OpenRouter API key** L83: - Sign up at https://openrouter.ai …
Critical
Ai Agent Control Hijack

Install-time source drops package-supplied AI-agent/MCP control files or instructions.

scripts/postinstall.jsView on unpkg · L5
src/sidecar/electron-install.jsView file
matchType = normalized_sha256 matchedPackage = amicus@1.9.1 matchedPath = src/sidecar/electron-install.js matchedIdentity = npm:YW1pY3Vz:1.9.1 similarity = 1.000 summary = normalized source hash matched finalized malicious source
High
Known Malware Source Similarity

Source file is highly similar to a previously finalized malicious package; route for source-aware review.

src/sidecar/electron-install.jsView on unpkg
src/sidecar/unzip.jsView file
matchType = normalized_sha256 matchedPackage = amicus@1.9.1 matchedPath = src/sidecar/unzip.js matchedIdentity = npm:YW1pY3Vz:1.9.1 similarity = 1.000 summary = normalized source hash matched finalized malicious source
High
Known Malware Source Similarity

Source file is highly similar to a previously finalized malicious package; route for source-aware review.

src/sidecar/unzip.jsView on unpkg
src/mcp-server.jsView file
matchType = previous_version_dangerous_delta matchedPackage = amicus@1.8.1 matchedIdentity = npm:YW1pY3Vz:1.8.1 similarity = 0.508 summary = stored previous version shares package body but lacks this dangerous source file
High
Previous Version Dangerous Delta

This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.

src/mcp-server.jsView on unpkg

Findings

1 Critical4 High5 Medium4 Low
CriticalAi Agent Control Hijackscripts/postinstall.js
HighInstall Time Lifecycle Scriptspackage.json
HighKnown Malware Source Similaritysrc/sidecar/electron-install.js
HighKnown Malware Source Similaritysrc/sidecar/unzip.js
HighPrevious Version Dangerous Deltasrc/mcp-server.js
MediumAmbiguous Install Lifecycle Scriptpackage.json
MediumDynamic Requirebin/amicus.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings