registry / AI-agent capability abuse / @p4code/cli / 0.4.14

@p4code/cli@0.4.14

P4 Code coding agent server and command-line client

AI Security Review

scanned 1mo ago · by lpm-firewall-ai

This is not malicious by static inspection: it has no install hook and its confirmed network request is a package update check. Its explicit agent-control and configuration-writing capability warrants a warning because it can alter a broad AI-agent control surface when invoked.

Static reason

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

Trigger

A user runs p4c and uses its MCP or server management features.

Impact

A configured MCP entry can affect later Claude sessions and may launch a user-selected command or contact a user-selected server.

Mechanism

Writes MCP server entries that may specify a command, environment, URL, or headers.

AI rationale

Evidence

dist/bin.mjs~/.claude.json<project>/.mcp.json

Decision evidence

public snapshot

AI called this Suspicious at 84.0% confidence as Dangerous Capability with medium false-positive risk.

Evidence for warning

The user-invoked CLI can write MCP registrations into the user-wide Claude configuration, which can configure commands or remote servers for future Claude sessions.
The bundled CLI exposes agent, MCP, skills, service, and synchronization commands, creating a broad agent-management capability.

Evidence against

package.json has no preinstall, install, or postinstall lifecycle hook.
The only confirmed automatic network request is an npm registry version check; it sends no environment or credential values.
The executable runs only when its p4c entry point is invoked.
No source evidence of credential exfiltration, obfuscated payload loading, or unsolicited destructive behavior was found.

Behavioral surface

Source

ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNativeBindingsNetworkShell

Supply chain

HighEntropyStrings

Source & flagged code

22 flagged

dist/client/assets/blade-llJRbbtR.jsView file

1import{t as e}from"./javascript-CUt1pgmJ.js";import{t}from"./css-DJp_X0uY.js";import{t as n}from"./html-C8UlPnhE.js";import{t as r}from"./xml-DIqSwXR3.js";import{t as i}from"./json... L2: //# sourceMappingURL=blade-llJRbbtR.js.map

High

Child Process

Package source references child process execution.

dist/client/assets/blade-llJRbbtR.jsView on unpkg · L1

dist/bin.mjsView file

13import * as NodeOS from "node:os"; L14: import * as NodeNet from "node:net"; L15: import * as Data from "effect/Data"; ... L21: import * as NodePath from "node:path"; L22: import * as NodeChildProcess from "node:child_process"; L23: import * as NodeFS from "node:fs"; ... L241: //#endregion L242: //#region package.json L243: var version = "0.4.14"; ... L376: /** L377: * Escapes a single argument for `cmd.exe` shell mode (`spawn(..., { shell: true })` L378: * on Windows). Node joins the command and arguments with spaces and hands the

Critical

Sensitive Archive Exfiltration

Source archives sensitive local configuration and publishes encoded chunks through an authenticated package-registry request.

dist/bin.mjsView on unpkg · L13

Critical

Credential Exfiltration

Source appears to send environment or credential material to an external endpoint.

dist/bin.mjsView on unpkg · L13

13Trigger-reachable chain: manifest.bin -> dist/bin.mjs L13: import * as NodeOS from "node:os"; L14: import * as NodeNet from "node:net"; L15: import * as Data from "effect/Data"; ... L21: import * as NodePath from "node:path"; L22: import * as NodeChildProcess from "node:child_process"; L23: import * as NodeFS from "node:fs"; ... L241: //#endregion L242: //#region package.json L243: var version = "0.4.14"; ... L376: /** L377: * Escapes a single argument for `cmd.exe` shell mode (`spawn(..., { shell: true })` L378: * on Windows). Node joins the command and arguments with spaces and hands the

Critical

Trigger Reachable Dangerous Capability

A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.

dist/bin.mjsView on unpkg · L13

364const POSIX_PATH_DELIMITER = ":"; L365: const WINDOWS_SHELL_CANDIDATES = ["pwsh.exe", "powershell.exe"]; L366: function canExecuteFile(filePath) {

High

Shell

Package source references shell execution.

dist/bin.mjsView on unpkg · L364

•Manifest-reachable source captures an API credential, sends it to a fixed unofficial gateway, and persists that redirection in agent or shell configuration. dist/bin.mjs: import * as NodeReadline from "node:readline"; }).pipe(Effect.flatMap((encoded) => fs.writeFileString(cachePath, encoded)), Effect.ignore); var EnvironmentAuthorizationError = class extends Schema$1.TaggedErrorClass()("EnvironmentAuthorizationError", { baseUrl: Schema$1.String authorization: Schema$1.optionalKey(Schema$1.String), httpBaseUrl: TrimmedNonEmptyString, wsBaseUrl: TrimmedNonEmptyString, httpBaseUrl: Schema$1.NullOr(Schema$1.String),

High

Credential Redirect Persistence

Manifest-reachable source sends a prompted API credential to a fixed unofficial gateway and persists the redirection.

dist/bin.mjsView on unpkg

High

Command Output Exfiltration

Source combines command execution, command-output handling, and outbound requests; review data flow before blocking.

dist/bin.mjsView on unpkg · L13

21import * as NodePath from "node:path"; L22: import * as NodeChildProcess from "node:child_process"; L23: import * as NodeFS from "node:fs"; ... L3738: name: TrimmedNonEmptyString, L3739: baseUrl: Schema$1.String L3740: });

High

Remote Agent Bridge

Source exposes local file and command tools to a remote model endpoint.

dist/bin.mjsView on unpkg · L21

13Trigger-reachable credential exfiltration chain: manifest.bin -> dist/bin.mjs L13: import * as NodeOS from "node:os"; L14: import * as NodeNet from "node:net"; L15: import * as Data from "effect/Data"; ... L21: import * as NodePath from "node:path"; L22: import * as NodeChildProcess from "node:child_process"; L23: import * as NodeFS from "node:fs"; ... L241: //#endregion L242: //#region package.json L243: var version = "0.4.14"; ... L376: /** L377: * Escapes a single argument for `cmd.exe` shell mode (`spawn(..., { shell: true })` L378: * on Windows). Node joins the command and arguments with spaces and hands the

High

Trigger Reachable Credential Exfiltration

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

dist/bin.mjsView on unpkg · L13

•stage = ast_semantic_analysis; reason = ast_path_work_budget_exceeded; limitedFiles = 9

High

Semantic Analysis Limited

A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.

dist/bin.mjsView on unpkg

•path = dist/bin.mjs kind = oversized_source_file sizeBytes = 5422291 magicHex = [redacted]

Medium

Oversized Source File

Package contains source files above the normal full-analysis size ceiling.

dist/bin.mjsView on unpkg

•path = dist/bin.mjs kind = oversized_cli_entrypoint sizeBytes = 5422291 magicHex = [redacted]

Medium

Oversized Cli Entrypoint

Package contains an oversized executable-looking CLI entrypoint.

dist/bin.mjsView on unpkg

dist/client/assets/sass-DXrisJhu.jsView file

1var e=[Object.freeze(JSON.parse(`{"displayName":"Sass","fileTypes":["sass"],"foldingStartMarker":"/\\\\*|^#|^\\\\*|^\\\\b|\\\\*#?region|^\\\\.","foldingStopMarker":"\\\\*/|\\\\*#?e... L2: //# sourceMappingURL=sass-DXrisJhu.js.map

Medium

Dynamic Require

Package source references dynamic require/import behavior.

dist/client/assets/sass-DXrisJhu.jsView on unpkg · L1

dist/NodePtyAdapter-BQAnH2v0.mjsView file

•matchType = normalized_sha256 matchedPackage = @p4code/cli@0.4.6 matchedPath = dist/NodePtyAdapter-BQAnH2v0.mjs matchedIdentity = npm:QHA0Y29kZS9jbGk:0.4.6 similarity = 1.000 summary = normalized source hash matched finalized malicious source

High