registry / AI-agent capability abuse / @tekmidian/pai / 0.64.0

@tekmidian/pai@0.64.0

PAI Knowledge OS — Personal AI Infrastructure with federated memory and project management

AI Security Review

scanned 11d ago · by lpm-firewall-ai

The source supports an explicit, consent-gated agent configuration capability, so this warrants a warning under the agent configuration policy. The scanner’s credential and gateway claims were not substantiated by the inspected excerpts.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

A user runs setup and accepts the settings patch prompt.

Impact

This enables package hooks in future Claude sessions; the inspected source did not establish credential theft or another active attack.

Mechanism

The setup flow registers package hooks and settings in the user’s Claude settings file.

AI rationale

Evidence

package.jsondist/program-JhF2GgJY.mjsdist/hooks/load-project-context.mjs~/.claude/settings.json

Network endpoints1

ntfy.sh

Decision evidence

public snapshot

AI called this Suspicious at 77.0% confidence as Dangerous Capability with medium false-positive risk.

Evidence for warning

The manifest has no preinstall, install, or postinstall hook; its prepublishOnly script runs the build and a worker command.
The setup command asks before adding hooks, environment variables, permissions, and flags to the user’s Claude settings.
After consent, setup registers package hooks in Claude settings; this is an explicit setup capability, not an install-time mutation.
The worker-provider example uses a placeholder endpoint and key-file path, and runs only after setup.

Evidence against

The flagged hook sends notifications to ntfy.sh only when the user supplies NTFY_TOPIC.
The reviewed sources did not establish credential forwarding to an unrelated recipient or an active attack.

Behavioral surface

Source

ChildProcessCryptoDynamicRequireEnvironmentVarsEvalFilesystemNetworkShell

Supply chain

HighEntropyStringsUrlStrings

Source & flagged code

13 flagged

dist/context-handover-cache-BOpjLsKe.mjsView file

2087} L2088: const { spawn } = await import("node:child_process"); L2089: return new Promise((resolve) => {

High

Child Process

Package source references child process execution.

dist/context-handover-cache-BOpjLsKe.mjsView on unpkg · L2087

dist/program-JhF2GgJY.mjsView file

10493const result = spawnSync(command, { L10494: shell: true, L10495: cwd,

High

Shell

Package source references shell execution.

dist/program-JhF2GgJY.mjsView on unpkg · L10493

•Manifest-reachable source captures an API credential, sends it to a fixed unofficial gateway, and persists that redirection in agent or shell configuration. dist/program-JhF2GgJY.mjs: import { appendFileSync, chmodSync, closeSync, copyFileSync, cpSync, createReadStream, existsSync, lstatSync, mkdirSync, openSync, readFileSync, readdirSync, readlinkSync, realpath... import { createInterface } from "node:readline"; * 08 - Others/MDF/Infrastruktur/.claude/worktrees/cool-haibt 1 session * 08 - Others/MDF/Infrastruktur/.claude/worktrees/strange-haibt 7 sessions fragment: `${sep}.claude${sep}worktrees${sep}`, writeFileSync(join(targetPath, "Notes", "MEMORY.md"), memoryContent, "utf-8"); const claudeProjects = join(homedir(), ".claude", "projects"); writeFileSync(configPath, JSON.strin

High

Credential Redirect Persistence

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

dist/program-JhF2GgJY.mjsView on unpkg

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. l.template.md"); if (!existssync(templatepath)) { console.log(c.warn("skill template not found: " + templatepath)); console.log(c.dim(" skipping pai skill installation.")); return false; } const skilldir = join(homedir(), ".claude", "skills", "pai"); const skillfile = join(skilldir, "skill.md"); if (existssync(skillfile)) { const content = readfilesync(skillfile, "utf-8"); const isgenerated = content.includes("generated by pai setup"); if (isgenerated) console.log(c.dim(" found existing pai-generated skill.md.")); else { console.log(c.yellow(" found existing skill.md (not pai-generated).")); console.log(c.dim(" a backup will be created before overwri

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/program-JhF2GgJY.mjsView on unpkg

32Cross-file remote execution chain: dist/program-JhF2GgJY.mjs spawns dist/daemon-mcp/index.mjs; helper contains network access plus dynamic code execution. L32: import { fileURLToPath } from "node:url"; L33: import { execFile, execFileSync, execSync, spawn, spawnSync } from "node:child_process"; L34: import { parseDocument } from "yaml"; ... L36: import { createInterface } from "node:readline"; L37: import { createConnection } from "net"; L38: import { getEncoding } from "js-tiktoken"; ... L52: * 08 - Others/MDF/Infrastruktur/.claude/worktrees/strange-haibt 7 sessions L53: * /private/tmp/ops-webui dead L54: * /private/tmp/claude-501/-Users-…-AIBroker/aae854c6-… dead ... L303: function findProjectNotesDirs(project) { L304: const cla

High

Cross File Remote Execution Context

Source spawns a local helper that also contains network and dynamic execution context; review data flow before blocking.

dist/program-JhF2GgJY.mjsView on unpkg · L32

32Trigger-reachable persistence chain: manifest.bin -> dist/cli/index.mjs -> dist/program-JhF2GgJY.mjs L32: import { fileURLToPath } from "node:url"; L33: import { execFile, execFileSync, execSync, spawn, spawnSync } from "node:child_process"; L34: import { parseDocument } from "yaml"; ... L36: import { createInterface } from "node:readline"; L37: import { createConnection } from "net"; L38: import { getEncoding } from "js-tiktoken"; ... L52: * 08 - Others/MDF/Infrastruktur/.claude/worktrees/strange-haibt 7 sessions L53: * /private/tmp/ops-webui dead L54: * /private/tmp/claude-501/-Users-…-AIBroker/aae854c6-… dead ... L303: function findProjectNotesDirs(project) { L304: const claudeProjects = join(homedir(), ".claude", "projects");

High

Trigger Reachable Persistence

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

dist/program-JhF2GgJY.mjsView on unpkg · L32

32import { fileURLToPath } from "node:url"; L33: import { execFile, execFileSync, execSync, spawn, spawnSync } from "node:child_process"; L34: import { parseDocument } from "yaml"; ... L36: import { createInterface } from "node:readline"; L37: import { createConnection } from "net"; L38: import { getEncoding } from "js-tiktoken"; ... L52: * 08 - Others/MDF/Infrastruktur/.claude/worktrees/strange-haibt 7 sessions L53: * /private/tmp/ops-webui dead L54: * /private/tmp/claude-501/-Users-…-AIBroker/aae854c6-… dead ... L303: function findProjectNotesDirs(project) { L304: const claudeProjects = join(homedir(), ".claude", "projects"); L305: if (!existsSync(claudeProjects)) return [];

Medium

Install Persistence

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

dist/program-JhF2GgJY.mjsView on unpkg · L32

dist/daemon-mcp/index.mjsView file

146try { L147: new Function(""); L148: return true;

High

Eval

Package source references dynamic code evaluation.

dist/daemon-mcp/index.mjsView on unpkg · L146

scripts/build-docs.mjsView file

335L336: const mod = await import(pathToFileURL(PROGRAM_MJS).href); L337: if (typeof mod.buildProgram !== "function") {

Medium

Dynamic Require

Package source references dynamic require/import behavior.

scripts/build-docs.mjsView on unpkg · L335

dist/hooks/load-project-context.mjsView file

3const require = __paiCreateRequire(import.meta.url); L4: if (!process.env.PAI_QUIET_NOTICES) process.env.PAI_QUIET_NOTICES = '1'; L5: var __create = Object.create; ... L2609: if (typeof node_buffer.Buffer === "function") { L2610: return node_buffer.Buffer.from(src, "base64"); L2611: } else if (typeof atob === "function") { ... L5223: yield* this.next(token); L5224: yield* this.end(forceDoc, endOffset); L5225: } ... L7429: import { join, dirname } from "node:path"; L7430: function paiHomeDir() { L7431: return process.env.PAI_HOME || join(homedir(), ".claude", "pai");

Critical

Credential Exfiltration

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

dist/hooks/load-project-context.mjsView on unpkg · L3

dist/hooks/cleanup-session-files.mjsView file

•Runtime or CLI code links a write operation to an explicit external AI-agent control path: import { existssync as existssync2, mkdirsync, readdirsync, linksync, copyfilesync } from "fs"; import { join as join2, basename } from "path"; import { homedir } from "os"; import { resolve, join } from "path"; import { existssync, readfilesync } from "fs"; function loadenvfile() { const possiblepaths = [ resolve(process.env.adapter_dir || process.env.pai_dir || "", ".env"), resolve(homedir(), ".claude", ".env") ]; for (const envpath of possiblepaths) { if (existssync(envpath)) { try { const content = readfilesync(envpath, "utf-8"); for (const line of content.split("\n")) { const trimmed = line.trim(); if (

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/hooks/cleanup-session-files.mjsView on unpkg

tab-color-command.shView file

•path = tab-color-command.sh kind = build_helper sizeBytes = 860 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

tab-color-command.shView on unpkg

dist/server-B71rem4q.mjs#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 2

High

Semantic Analysis Limited

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

dist/server-B71rem4q.mjs#virtual:normalized:round1View on unpkg

Source & flagged code

13 flagged · loading source

dist/context-handover-cache-BOpjLsKe.mjsView file

2087} L2088: const { spawn } = await import("node:child_process"); L2089: return new Promise((resolve) => {

High

Child Process

Package source references child process execution.

dist/context-handover-cache-BOpjLsKe.mjsView on unpkg · L2087

dist/program-JhF2GgJY.mjsView file

10493const result = spawnSync(command, { L10494: shell: true, L10495: cwd,

High

Shell

Package source references shell execution.

dist/program-JhF2GgJY.mjsView on unpkg · L10493

Findings

1 Critical9 High6 Medium5 Low

CriticalCredential Exfiltrationdist/hooks/load-project-context.mjs

HighChild Processdist/context-handover-cache-BOpjLsKe.mjs

HighShelldist/program-JhF2GgJY.mjs

HighEvaldist/daemon-mcp/index.mjs

HighCredential Redirect Persistencedist/program-JhF2GgJY.mjs

HighExternal Ai Agent Control Surface Mutationdist/hooks/cleanup-session-files.mjs

HighTrigger Reachable External Ai Agent Control Surface Mutationdist/program-JhF2GgJY.mjs

HighCross File Remote Execution Contextdist/program-JhF2GgJY.mjs

HighTrigger Reachable Persistencedist/program-JhF2GgJY.mjs

HighSemantic Analysis Limiteddist/server-B71rem4q.mjs#virtual:normalized:round1

Affected versions and remediation

This report applies to @tekmidian/pai@0.64.0.

See all versions of @tekmidian/pai for other recorded verdicts.

Review the evidence and your use of @tekmidian/pai@0.64.0 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: 2026-10-11 18:32:20.382Z (UTC).

Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Affected versions and remediation

Runtime surface

Behavioral surface

Source & flagged code