@whalent/agent-core@0.3.470
Core runtime for Whalent Agent
AI Security Review
scanned 26d ago · by lpm-firewall-aiThe bundle can rewrite foreign agent MCP configs and skill directories, but only through an explicit daemon tool action, and package.json has no install hook. That is a guarded agent-control capability, so it warrants a warning rather than a publish block.
Decision evidence
public snapshot- A daemon tool action named register calls injectMcpServer with the caller-supplied server name, command, args, and env.
- injectMcpServer writes that server into Claude, OpenCode, Kimi, or Codex config, including ~/.claude.json via writeClaudeConfig.
- installSkill creates a directory from platformDir (.claude, .cursor, or .agents) under the home directory or workspace and writes skill files there.
- package.json exposes the whalent-core CLI and has no install lifecycle scripts, so these writes run only when the daemon tool is invoked.
- Published package.json has no preinstall, install, or postinstall hook.
- Skill file writes reject absolute paths and paths that escape the skill directory.
- The platform token is a caller-supplied Whalent token sent to the configured Whalent platform, defaulting to memory.whalent.com.
Source & flagged code
8 flagged · loading sourceA single source file combines environment access, network access, and code or shell execution; review context before blocking.
dist/mcp.cjsView on unpkg · L4Source contains an obfuscated payload loader that reconstructs and executes hidden code.
dist/mcp.cjsView on unpkg · L1Findings
Affected versions and remediation
This report applies to @whalent/agent-core@0.3.470.
See all versions of @whalent/agent-core for other recorded verdicts.
- Review the evidence and your use of @whalent/agent-core@0.3.470 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.