Loading npm security reports…

greprag: AI-agent control-surface risk | LPM Firewall

registry / greprag / 5.78.1

greprag@5.78.1

GrepRAG — agent memory for Claude Code, Codex, OpenCode, and Grok Build.

AI Security Review

scanned 2d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. Installing the package runs a postinstall script that mutates pre-existing Claude Code and Codex skill directories without an explicit setup command. The installed agent skill can subsequently direct agent behavior; the package also has an explicit runtime turn-capture uploader.

Static reason

High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review

Trigger

npm install where ~/.claude or ~/.codex already exists

Impact

Unconsented persistent agent control-surface modification; later runtime capture can transmit conversation content.

Mechanism

postinstall copies package-owned skill content into foreign AI-agent skill registries

Policy narrative

On installation, the declared postinstall scans for existing Claude Code and Codex homes and copies the package’s core skill into their skill registries. This occurs without the user invoking GrepRAG setup or approving the injected skill. The package’s watcher separately captures OpenCode user prompts and assistant responses and posts them to its API when the watcher is run.

Rationale

This is a concrete unconsented postinstall mutation of foreign AI-agent control surfaces, meeting the blocking policy. Explicit later setup and trust-review guidance does not mitigate the earlier automatic skill installation.

Evidence

package.jsonscripts/postinstall.jsskill/greprag/SKILL.mddist/commands/opencode-watch.js~/.claude/skills/greprag~/.codex/skills/greprag

Network endpoints1

Source & flagged code

13 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

dist/commands/grok-spawn.jsView file

5Trigger-reachable persistence chain: manifest.main -> dist/index.js -> dist/commands/grok-spawn.js L5: * children when the tool call returns. The surviving launch is a scheduled L6: * task (`schtasks /Create /IT` then `/Run`) that `start`s a new `cmd /k` L7: * window running grok.exe — a visible command-prompt bootloader outside the ... L50: const path = __importStar(require("path")); L51: const child_process_1 = require("child_process"); L52: function winQuote(value) { ... L55: function resolveGrokExe() { L56: const home = os.homedir(); L57: const named = process.platform === 'win32' ? 'grok.exe' : 'grok'; L58: const pinned = path.join(home, '.grok', 'bin', named); ... L136: console.error(` ${cmd.map(a => /\s/.test(a) ? JSON.stringify(a) : a).join(' ')}`); L137: process.exitCode = 1;

High

Trigger Reachable Persistence

A manifest entrypoint or package-local install chain reaches persistence behavior.

dist/commands/grok-spawn.jsView on unpkg · L5

5* children when the tool call returns. The surviving launch is a scheduled L6: * task (`schtasks /Create /IT` then `/Run`) that `start`s a new `cmd /k` L7: * window running grok.exe — a visible command-prompt bootloader outside the ... L50: const path = __importStar(require("path")); L51: const child_process_1 = require("child_process"); L52: function winQuote(value) { ... L55: function resolveGrokExe() { L56: const home = os.homedir(); L57: const named = process.platform === 'win32' ? 'grok.exe' : 'grok'; L58: const pinned = path.join(home, '.grok', 'bin', named); ... L136: console.error(` ${cmd.map(a => /\s/.test(a) ? JSON.stringify(a) : a).join(' ')}`); L137: process.exitCode = 1;

Medium

Install Persistence

Source writes installer persistence such as shell profile or service configuration.

dist/commands/grok-spawn.jsView on unpkg · L5

dist/commands/opencode-watch.jsView file

•Manifest-reachable source resolves another installed package, overwrites its runtime code, and injects package-defined remote behavior. dist/commands/opencode-watch.js: const API_URL_DEFAULT = 'https://api.greprag.com'; fs.writeFileSync(LOCK_FILE, String(process.pid)); fs.writeFileSync(file, String(timeMs)); locateFile: (f) => path.join(path.dirname(require.resolve('sql.js')), f),

High

Entrypoint Foreign Package Code Overwrite

Manifest-reachable source overwrites another installed package with package-defined remote behavior.

dist/commands/opencode-watch.jsView on unpkg

•matchType = normalized_sha256 matchedPackage = greprag@5.77.0 matchedPath = dist/commands/opencode-watch.js matchedIdentity = npm:Z3JlcHJhZw:5.77.0 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.

dist/commands/opencode-watch.jsView on unpkg

scripts/postinstall.jsView file

24name: 'Claude Code', L25: agentDir: path.join(home, '.claude'), L26: skillsTarget: path.join(home, '.claude', 'skills'), L27: docsTarget: path.join(home, '.claude', 'docs'), L28: refreshClaudeDocs: true, ... L31: name: 'Codex', L32: agentDir: path.join(home, '.codex'), L33: skillsTarget: path.join(home, '.codex', 'skills'), L34: docsTarget: path.join(home, '.codex', 'docs'), L35: refreshClaudeDocs: false, ... L39: function copyDir(src, dest) { L40: fs.mkdirSync(dest, { recursive: true });

Critical

Ai Agent Control Hijack

Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.

scripts/postinstall.jsView on unpkg · L24

dist/opencode-plugin.bundle.jsView file

•Runtime or CLI code links a write operation to an explicit external AI-agent control path: tivation(entries, readnativeskillnames(params)); } // src/opencode-plugin.ts var debug_log_path = path8.join(os3.homedir(), ".greprag", "opencode-plugin-debug.log"); var _debuglogready = false; function dloginit() { if (_debuglogready) return; _debuglogready = true; try { fs8.writefilesync(debug_log_path, ""); } catch { } } function dlog(msg) { dloginit(); const line = `[${(/* @__pure__ */ new date()).toisostring()}] [greprag-memory] ${msg} `; try { fs8.appendfilesync(debug_log_path, line); } catch { } } dlog(`module top reached: pid=${process.pid} argv0=${process.argv[0]} distpath=${__filename}`); var api_url = "https: function loadenvfile(filepath

High

External Ai Agent Control Surface Mutation

Runtime or CLI source writes behavior-bearing configuration into a user or project AI-agent control surface.

dist/opencode-plugin.bundle.jsView on unpkg

•matchType = normalized_sha256 matchedPackage = greprag@5.77.0 matchedPath = dist/opencode-plugin.bundle.js matchedIdentity = npm:Z3JlcHJhZw:5.77.0 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.

dist/opencode-plugin.bundle.jsView on unpkg

dist/commands/init.jsView file

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. ) fs.mkdirsync(dir, { recursive: true }); const body = object.entries(existing) .map(([k, v]) => `${k}=${v}`) .join('\n') + '\n'; fs.writefilesync(file, body); } function readenvfile(file) { const out = {}; try { const raw = fs.readfilesync(file, 'utf-8'); for (const line of raw.split(/\r?\n/)) { const trimmed = line.trim(); if (!trimmed || trimmed.startswith('#')) continue; const idx = trimmed.indexof('='); if (idx < 1) continue; let value = trimmed.slice(idx + 1).trim(); if ((value.startswith('"') && value.endswith('"')

High

Trigger Reachable External Ai Agent Control Surface Mutation

Manifest-trigger-reachable source writes behavior-bearing configuration into a user or project AI-agent control surface.

dist/commands/init.jsView on unpkg

dist/commands/watcher-registry.jsView file

•matchType = normalized_sha256 matchedPackage = greprag@5.77.0 matchedPath = dist/commands/watcher-registry.js matchedIdentity = npm:Z3JlcHJhZw:5.77.0 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.

dist/commands/watcher-registry.jsView on unpkg

dist/ccr-store.jsView file

•matchType = normalized_sha256 matchedPackage = greprag@5.77.0 matchedPath = dist/ccr-store.js matchedIdentity = npm:Z3JlcHJhZw:5.77.0 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.

dist/ccr-store.jsView on unpkg

dist/commands/doctor.jsView file

•matchType = normalized_sha256 matchedPackage = greprag@5.77.0 matchedPath = dist/commands/doctor.js matchedIdentity = npm:Z3JlcHJhZw:5.77.0 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.

dist/commands/doctor.jsView on unpkg