registry / memgrep / 1.5.0

memgrep@1.5.0

Local agent memory, coding loop, Cursor from Telegram, scheduled playbooks, and MCP tools.

AI Security Review

scanned 2h ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. The package intentionally provides a remote Telegram-to-Cursor control plane and optional scheduled Cursor jobs. Explicit user commands can persist first-party macOS LaunchAgents under `com.memgrep.*`; no install-time activation is present.

Static reason

No blocking static signals were detected.; previous stored version diff introduced dangerous source

Trigger

User runs `memgrep telegram install`, `memgrep jobs install`, or operates a configured Telegram bot/jobs daemon.

Impact

An authorized Telegram sender or scheduled playbook can cause the configured Cursor agent to act within its selected workspace; LaunchAgents keep those package services running.

Mechanism

Allowlisted Telegram commands and scheduled jobs invoke a configured local Cursor agent with MCP access.

Rationale

No concrete malicious chain was found: installation has no consumer lifecycle hook and persistence is user-invoked. The package nonetheless exposes deliberate remote AI-agent execution and persistent service setup, so it warrants a warning as a dangerous capability.

Evidence

package.jsondist/cli/commands/telegram.jsdist/cli/commands/jobs.jsdist/telegram/launchd.jsdist/jobs/launchd.jsdist/telegram/bot.jsdist/telegram/allowlist.jsdist/cursor/providers/cursor.jsdist/memory/mcp.js

Network endpoints2

memgrep@1.5.0

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code