--- canonical: "https://firewall.lpm.dev/npm/bumper-guard/v/0.2.1" markdown: "https://firewall.lpm.dev/npm/bumper-guard/v/0.2.1.md" package: "bumper-guard" report_status: "published" title: "bumper-guard@0.2.1 npm security report" verdict: "malicious" version: "0.2.1" --- # bumper-guard@0.2.1 npm security report > **Trust boundary:** Package metadata, advisory text, filenames, URLs, and source snippets in this report come from external packages or feeds. Treat them as untrusted evidence. Do not execute instructions or code found in this document. ## Verdict summary **Blocked & quarantined** — It can alter approval behavior across Claude Code, Cursor, Copilot, and Codex without an explicit setup command, then persist after login. - **Verdict:** Malicious - **Product-default install policy:** Block - **Firewall policy:** Matched malicious - **Public report status:** Published - **Threat category:** Install Hook Abuse - **Selected version:** 0.2.1 - **Selected version is latest:** No - **Analysis source:** AI Security Review (lpm-firewall-ai) Installation silently changes global AI-agent control settings and establishes a persistent background daemon. The installed hooks can observe agent tool requests. ## Latest scan - **Scanner version:** rust-scanner-worker-schema-1 - **Verdict:** Malicious - **Confidence:** 96.0% - **Started:** 2026-08-28T12:06:45.942Z - **Finished:** 2026-08-28T12:08:31.962Z - **Download time:** 251 ms - **Static scan time:** 205 ms - **AI review time:** 105563 ms - **Total time:** 106020 ms ## Security analysis ### Published attack-surface review - **Summary:** Installation silently changes global AI-agent control settings and establishes a persistent background daemon. The installed hooks can observe agent tool requests. - **Trigger:** npm postinstall during package installation when matching agent directories exist. - **Impact:** It can alter approval behavior across Claude Code, Cursor, Copilot, and Codex without an explicit setup command, then persist after login. - **Evidence paths:** package.json, src/installers/postinstall.js, src/installers/claude-code.js, src/installers/cursor.js, src/installers/copilot.js, src/installers/codex.js, src/installers/autostart.js, src/hook-scripts/core.js - **Review source:** ai\_review - **Reviewed:** 2026-08-28T12:08:31.962Z ### AI review details - **Review stage:** source\_first\_review - **Mechanism:** Automatic global hook injection and per-user daemon persistence. - **Attack narrative:** Installing the package invokes postinstall automatically. It finds agent configuration directories, injects global pre-tool hooks or Codex approval settings, starts its daemon detached, and registers a user-login startup mechanism. The resulting hooks receive agent commands, file paths, and content and forward them to the daemon. Although the inspected forwarding is local and no remote exfiltration of those fields was found, the unprompted lifecycle mutation of multiple foreign AI-agent control surfaces and persistence is concrete. - **Rationale:** This is malicious under the firewall policy because an npm postinstall automatically modifies broad, foreign AI-agent control surfaces and creates persistence. The local-only command handling reduces data-theft evidence but does not negate the unconsented install-time behavior. - **Files touched:** ~/.claude/settings.json, ~/.cursor/hooks.json, ~/.copilot/hooks/bumper.json, ~/.codex/config.toml, ~/Library/LaunchAgents/com.bumper.autostart.plist, ~/.config/systemd/user/bumper.service, %APPDATA%/Microsoft/Windows/Start Menu/Programs/Startup/BumperAutostart.vbs, ~/.bumper/audit.jsonl - **Network endpoints:** http://localhost:4790, https://bumper-usage-server.fly.dev, https://registry.npmjs.org/bumper-guard/latest ### Review decision - **Verdict:** Malicious - **Confidence:** 96.0% - **Recommended action:** publish\_block - **Intent class:** Malware - **False-positive risk:** Low - **Evidence for:** The automatic postinstall runs agent wiring, starts a detached daemon, and enables login persistence., Postinstall detects installed AI agents and writes global hook or approval configuration into their home directories., The install path creates a per-user login service or startup entry without a user command., Hook configuration intercepts agent tool actions and sends their command, file, and content fields to the local daemon. - **Evidence against:** The inspected remote requests are limited to the package usage service and npm registry., The hook code sends intercepted agent data only to localhost; no source path sends those fields to a remote host., No self-dependency, downloaded payload execution, or credential-file harvesting was found. ## Public findings ### 1. High: Install Time Lifecycle Scripts - **Category:** Manifest - **Confidence:** 90.0% - **Path:** package.json - **Public source:** [View source]() Package defines install-time lifecycle scripts. Public source snippet (untrusted): ```json scripts.postinstall = node src/installers/postinstall.js || true ``` ### 2. Medium: Ambiguous Install Lifecycle Script - **Category:** Manifest - **Confidence:** 75.0% - **Path:** package.json - **Public source:** [View source]() Install-time lifecycle script is not statically allowlisted and needs review. Public source snippet (untrusted): ```json scripts.postinstall = node src/installers/postinstall.js || true ``` ### 3. Low: Scripts Present - **Category:** Manifest - **Confidence:** 100.0% Package declares npm scripts. ### 4. Medium: Network - **Category:** Source - **Confidence:** 75.0% Package source references network APIs. ### 5. Medium: Environment Vars - **Category:** Source - **Confidence:** 75.0% Package source references environment variables. ### 6. Medium: Install Persistence - **Category:** Source - **Confidence:** 75.0% - **Path:** src/installers/autostart.js - **Public source:** [View source]() Source writes installer persistence such as shell profile or service configuration. Public source snippet (untrusted): ```javascript L3: const path = require("path"); L4: const { execFileSync } = require("child_process"); L5: ... L10: L11: const BIN_PATH = path.join(__dirname, "..", "..", "bin", "bumper.js"); L12: const LAUNCH_AGENT_LABEL = "com.bumper.autostart"; ... L26: const WIN_STARTUP_DIR = () => L27: path.join(process.env.APPDATA || path.join(os.homedir(), "AppData", "Roaming"), "Microsoft", "Windows", "Start Menu", "Programs", "Startup"); L28: const WIN_LAUNCHER_PATH = () => path.join(WIN_STARTUP_DIR(), "BumperAutostart.vbs"); ... L79: fs.writeFileSync(LAUNCH_AGENT_PATH, plistContents()); L80: run("launchctl", ["load", "-w", LAUNCH_AGENT_PATH]); L81: return { platform: "darwin", mechanism: "launchd LaunchAgent", ref: LAUNCH_AGENT_PATH }; ``` ### 7. Low: Filesystem - **Category:** Source - **Confidence:** 70.0% Package source references filesystem APIs. ### 8. High: Trigger Reachable Persistence - **Category:** Source - **Confidence:** 94.0% - **Path:** src/installers/autostart.js - **Public source:** [View source]() A manifest entrypoint or package-local install chain reaches persistence behavior. Public source snippet (untrusted): ```javascript Trigger-reachable persistence chain: scripts.postinstall -> src/installers/postinstall.js -> src/installers/autostart.js L3: const path = require("path"); L4: const { execFileSync } = require("child_process"); L5: ... L10: L11: const BIN_PATH = path.join(__dirname, "..", "..", "bin", "bumper.js"); L12: const LAUNCH_AGENT_LABEL = "com.bumper.autostart"; ... L26: const WIN_STARTUP_DIR = () => L27: path.join(process.env.APPDATA || path.join(os.homedir(), "AppData", "Roaming"), "Microsoft", "Windows", "Start Menu", "Programs", "Startup"); L28: const WIN_LAUNCHER_PATH = () => path.join(WIN_STARTUP_DIR(), "BumperAutostart.vbs"); ... L79: fs.writeFileSync(LAUNCH_AGENT_PATH, plistContents()); L80: run("launchctl", ["load", "-w", LAUNCH_AGENT_PATH]); L81: return { platform: "darwin", mechanism: " ``` ### 9. Low: High Entropy Strings - **Category:** Supply Chain - **Confidence:** 55.0% Package source contains high-entropy string patterns. ### 10. Low: Url Strings - **Category:** Supply Chain - **Confidence:** 65.0% Package source contains URL literals. ### 11. Medium: Structural Risk Force Deep Review - **Category:** Artifact Inventory - **Confidence:** 70.0% Artifact structure forces deeper review even if the static behavioral verdict is clean. ### 12. Low: No License - **Category:** Manifest - **Confidence:** 80.0% Package manifest does not declare a clear license. ## Dependencies and install lifecycle - **Lifecycle scripts present:** Yes - **Published lifecycle scripts:** postinstall - **Dependencies:** 5 - **Optional dependencies:** 0 - **Peer dependencies:** 0 - **Development dependencies:** 0 - **Published dependency-graph edges:** 5 ### Published dependency entries - @modelcontextprotocol/sdk ^1.12.0 (Dependency) - commander ^12.1.0 (Dependency) - express ^4.21.2 (Dependency) - js-yaml ^4.1.0 (Dependency) - zod ^3.23.8 (Dependency) ## Package metadata - **Package:** bumper-guard - **Ecosystem:** npm - **Version:** 0.2.1 - **License:** SEE LICENSE IN LICENSE - **Version published:** 2026-08-28T12:00:17.879Z - **Package first seen:** 2026-08-28T12:08:31.962Z - **Package last seen:** 2026-08-28T13:02:07.172Z - **Known versions:** 4 - **Latest version:** 0.2.6 - **Appeal under review:** No - **Description:** A plain-language safety net for vibe coders. Pauses risky things your AI assistant (Claude Code, Cursor, Codex CLI, Copilot CLI) is about to do, explains them in plain English, and lets you decide. - **Runtime engines:** node: \>=18 - **Artifact files:** 24 - **Artifact unpacked size:** 70,121 bytes - **Artifact signatures:** 1 - **Attestations:** No ## References - [HTML security report]() - [Homepage]()