registry / @whalent/agent-core / 0.3.468
@whalent/agent-core@0.3.468
Core runtime for Whalent Agent
AI Security Review
scanned 27d ago · by lpm-firewall-aiThe bundle can mutate other agents' skill and MCP configuration, which is a real control-surface capability, but package.json has no install hook and the MCP daemon requires an explicit token argument. Obfuscation hides the exact consent gate, so the behavior is a guarded first-party setup risk rather than a proven install-time hijack.
Static reason
One or more suspicious static signals were detected.
Trigger
A user or parent wrapper runs the whalent-core CLI or the bundled MCP entrypoint with arguments that reach skill installation or MCP registration.
Impact
An invoked setup path can change which commands Claude, Cursor, or similar local agents launch. No install-time or silent exfiltration chain was established.
Mechanism
platformDir chooses a home or workspace agent directory. writeSkillFiles then writes skill files there. injectClaudeMcpServer and sibling helpers merge MCP server command entries into existing agent config JSON and save the result.
AI rationale
The bundle can mutate other agents' skill and MCP configuration, which is a real control-surface capability, but package.json has no install hook and the MCP daemon requires an explicit token argument. Obfuscation hides the exact consent gate, so the behavior is a guarded first-party setup risk rather than a proven install-time hijack.
Evidence
dist/mcp.cjs.claude.jsonmcp.json.cursor
Decision evidence
public snapshotAI called this Suspicious at 74.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- package.json has no install lifecycle scripts. The bin command whalent-core points at dist/index.cjs, and dependencies are only ssh2 plus optional node-pty.
- dist/mcp.cjs maps Claude and cursor_cli platforms onto home agent directories, including a literal .cursor directory.
- writeSkillFiles writes supplied skill file contents under those resolved directories after rejecting absolute paths and paths that escape the skill root.
- injectClaudeMcpServer reads .claude.json, inserts an mcpServers entry, and writes the file back. Similar helpers exist for other agent MCP configs such as mcp.json.
- The mcp.cjs entrypoint only starts its daemon after command-line arguments supply a token. It is not wired to an npm install hook.
Evidence against
- No preinstall, install, or postinstall script is declared, and the package does not depend on its own name.
- The daemon startup path exits when the token argument is missing.
- Skill file writes are confined by a path check before writeFile.
- The only clearly decoded service host is an origin fallback limited to memory.whalent.com and numbered api hosts under that domain.
Source & flagged code
7 flagged · loading sourcedist/mcp.cjsView file
1#!/usr/bin/env node
L2: 'use strict';var a0_0x21d998=a0_0x1d06;(function(_0x342ea1,_0x584be1){var _0x530453=a0_0x1d06,_0x428a3b=_0x342ea1();while(!![]){try{var _0xd872c4=parseInt(_0x530453(0x20f))/0x1+par...
L3: || (${_0x4b80f3} == "string" && ${_0x10a25c} && ${_0x10a25c} == +${_0x10a25c})`)[_0x2297e4(0x1c18)](_0x47b462,(0x0,_0x541a1c['_'])`+${_0x10a25c}`);return;case'integer':_0x3c8cb0['e...
High
Obfuscated Payload Loader
Source contains an obfuscated payload loader that reconstructs and executes hidden code.
dist/mcp.cjsView on unpkg · L1•stage = ast_semantic_analysis; reason = ast_path_work_budget_exceeded; limitedFiles = 1
High
Semantic Analysis Limited
A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.
dist/mcp.cjsView on unpkg4|| (${_0x4b80f3} === "string" && ${_0x10a25c} && ${_0x10a25c} == +${_0x10a25c} && !(${_0x10a25c} % 1))`)[_0x2297e4(0x1c18)](_0x47b462,(0x0,_0x541a1c['_'])`+${_0x10a25c}`);return;ca...
L5: || ${_0x4b80f3} === "boolean" || ${_0x10a25c} === null`)['assign'](_0x47b462,(0x0,_0x541a1c['_'])`[${_0x10a25c}]`);}}}function _0x3084bd({gen:_0x15c7d9,parentData:_0x11808a,parentD...
L6: missingProperty: ${_0x27b044},
Medium
Findings
3 High7 Medium5 Low
HighObfuscated Payload Loaderdist/mcp.cjs
HighObfuscated
HighSemantic Analysis Limiteddist/mcp.cjs
MediumDynamic Requiredist/mcp.cjs
MediumNetwork
MediumEnvironment Vars
MediumShips Build Helperdist/jupyter_sidecar.py
MediumOversized Source Filedist/index.cjs
MediumOversized Cli Entrypointdist/index.cjs
MediumStructural Risk Force Deep Review
LowEvaldist/mcp.cjs
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
LowNo License
Affected versions and remediation
This report applies to @whalent/agent-core@0.3.468.
See all versions of @whalent/agent-core for other recorded verdicts.
- Review the evidence and your use of @whalent/agent-core@0.3.468 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.