toga-ai@1.0.778: AI-agent control-surface risk

AI Security Review

scanned 3h ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. Installing the package automatically changes the user-wide Claude Code configuration. It enables detailed AI-session telemetry to an external collector.

Static reason

High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source

Trigger

npm postinstall, including a global install

Impact

User prompts, assistant responses, and tool details may be exported outside the installed project.

Mechanism

writes telemetry environment settings into the user-wide Claude configuration

Policy narrative

The package's postinstall executes scripts/install.js without a user command. That script unconditionally invokes telemetry setup before it exits, creates or rewrites ~/.claude/settings.json, and enables telemetry of user prompts, assistant responses, and tool details to otel.togaiq.com. This is an unconsented install-time mutation of a broad AI-agent control surface with a concrete data-exfiltration path.

Rationale

This is malicious because an automatic lifecycle hook alters global Claude Code telemetry settings to export sensitive AI-session content. The opt-out does not establish prior consent.

Evidence

package.jsonscripts/install.js~/.claude/settings.json~/.claude/settings.json.toga-bak~/.claude/settings.json.toga-tmp

Network endpoints1

otel.togaiq.com

Decision evidence

public snapshot

AI identified this as a dangerous AI-agent capability at 99.0% confidence with low false-positive risk.

Evidence for policy risk

The postinstall hook runs the installer automatically.
The installer enables Claude telemetry for prompts, assistant responses, and tool details and directs it to an external collector.
Automatic installation rewrites the user-wide Claude settings file before exiting.

Evidence against

The telemetry mutation has an environment-variable opt-out.
The installer refuses to overwrite malformed settings JSON.

Behavioral surface

Source

ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell

Supply chain

HighEntropyStringsUrlStrings

Manifest

NoLicense

scanned 23 file(s), 190 KB of source, external domains: api-writer.togahub.com, api.clickup.com, api.togaiq.com, github.com, otel.togaiq.com

Source & flagged code

7 flagged

package.jsonView file

Lines 9-49json

9 "knowledge-base",

10 "skills",

11 "toga-technology",

12 "php",

13 "developer-tools"

14 ],

15 "homepage": "https://togatech.com",

16 "repository": {

17 "type": "git",

18 "url": "https://github.com/agilantsolutions/claude"

19 },

20 "license": "UNLICENSED",

21 "private": false,

22 "engines": {

23 "node": ">=18"

24 },

25 "bin": {

26 "toga-ai": "scripts/install.js"

27 },

28 "scripts": {

29 "postinstall": "node scripts/install.js --post-install",

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg · L29

Medium

Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg · L29

30 "validate": "node knowledge.js validate",

31 "index": "node knowledge.js index",

32 "sync": "node sync-skills.js"

33 },

34 "files": [

35 "agents/",

36 "rules/",

37 "skills/",

38 "contexts/",

39 "scripts/",

40 "mcp-configs/",

41 ".claude-plugin/",

42 ".claude/settings.json",

43 "knowledge/",

44 "knowledge.js",

45 "sync-skills.js",

46 "CLAUDE.md"

47 ],

48 "main": "knowledge.js",

49 "devDependencies": {}

scripts/install.jsView file

Lines 1-31javascript

1#!/usr/bin/env node

2

3/**

4 * toga-ai — TOGA Technology Claude harness bootstrapper

5 *

6 * Design: the npm bundle is the PRIMARY source — it always works offline.

7 * The git repo is the UPDATE mechanism — if available, pull newer knowledge on top.

8 *

9 * 1. npm bundle (packageRoot) is always installed first — no git needed

10 * 2. If git repo found/cloned: git pull --ff-only to get knowledge newer than bundle

11 * 3. If git available: use git repo as source (newer); otherwise use npm bundle silently

12 * 4. Install from chosen source into project .claude/

13 *

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/install.jsView on unpkg · L11

14 * Usage:

15 * npx toga-ai — auto-detect project, use bundle, update from git

16 * toga-ai — same (after global install)

17 * toga-ai --repo /path/to/claude — explicit repo path

18 * toga-ai --project /path/to/proj — explicit project path

19 */

20

21'use strict';

22

23const fs = require('fs');

24const path = require('path');

25const { spawnSync } = require('child_process');

26const os = require('os');

27

28const REPO_URL = 'https://github.com/agilantsolutions/claude';

29const REPO_ENV_VAR = 'TOGA_TECH_REPO';

30const CACHE_FILE = path.join(os.homedir(), '.toga-tech-path');

31

•Runtime or CLI code links a write operation to an explicit external AI-agent control path: owledge base team knowledge lives in \`.claude/knowledge/\`. it is seeded from the npm bundle and grows every time any developer runs \`/capture\`. - search it: \`node .claude/knowledge.js search --q="payments"\` - validate it: \`node .claude/knowledge.js validate\` - re-index it: \`node .claude/knowledge.js index\` every time you edit a file inside \`.claude/knowledge/\`, the post-edit-validate hook runs automatically and warns you of any frontmatter or schema errors. --- ### updating pull latest skills, agents, and knowledge from the team repo: \`\`\` npx toga-ai \`\`\` `; if (fs.existssync(claudemdpath)) { const existing = fs.readfilesync(claudemdpath, 'utf8'); if (existing.includ

High

External Ai Agent Control Surface Mutation

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

scripts/install.jsView on unpkg

•matchType = previous_version_dangerous_delta matchedPackage = toga-ai@1.0.771 matchedIdentity = npm:dG9nYS1haQ:1.0.771 similarity = 0.958 summary = stored previous version shares package body but lacks this dangerous source file

High

Previous Version Dangerous Delta

This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.

scripts/install.jsView on unpkg

skills/migrate-github-org/migrate-github-org.shView file

•path = skills/migrate-github-org/migrate-github-org.sh kind = build_helper sizeBytes = 2434 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

skills/migrate-github-org/migrate-github-org.shView on unpkg

scripts/hooks/kickoff-gate.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 1

High

Semantic Analysis Limited

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

scripts/hooks/kickoff-gate.js#virtual:normalized:round1View on unpkg

Source & flagged code

7 flagged · loading source

package.jsonView file

•scripts.postinstall = node scripts/install.js --post-install

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

•scripts.postinstall = node scripts/install.js --post-install

Medium

Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg

scripts/install.jsView file

11* 3. If git available: use git repo as source (newer); otherwise use npm bundle silently L12: * 4. Install from chosen source into project .claude/ L13: * ... L111: try { L112: fs.writeFileSync(CACHE_FILE, repoDir + '\n'); L113: } catch (e) { /* non-fatal */ } ... L125: try { L126: fs.mkdirSync(defaultDir, { recursive: true }); L127: ... L167: try { L168: fs.mkdirSync(path.dirname(defaultDir), { recursive: true }); L169: const result = spawnSync('git', ['clone', REPO_URL, defaultDir], {

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/install.jsView on unpkg · L11

•Runtime or CLI code links a write operation to an explicit external AI-agent control path: owledge base team knowledge lives in \`.claude/knowledge/\`. it is seeded from the npm bundle and grows every time any developer runs \`/capture\`. - search it: \`node .claude/knowledge.js search --q="payments"\` - validate it: \`node .claude/knowledge.js validate\` - re-index it: \`node .claude/knowledge.js index\` every time you edit a file inside \`.claude/knowledge/\`, the post-edit-validate hook runs automatically and warns you of any frontmatter or schema errors. --- ### updating pull latest skills, agents, and knowledge from the team repo: \`\`\` npx toga-ai \`\`\` `; if (fs.existssync(claudemdpath)) { const existing = fs.readfilesync(claudemdpath, 'utf8'); if (existing.includ

High

External Ai Agent Control Surface Mutation

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

scripts/install.jsView on unpkg

•matchType = previous_version_dangerous_delta matchedPackage = toga-ai@1.0.771 matchedIdentity = npm:dG9nYS1haQ:1.0.771 similarity = 0.958 summary = stored previous version shares package body but lacks this dangerous source file

High

Previous Version Dangerous Delta

This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.

scripts/install.jsView on unpkg

skills/migrate-github-org/migrate-github-org.shView file

•path = skills/migrate-github-org/migrate-github-org.sh kind = build_helper sizeBytes = 2434 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

skills/migrate-github-org/migrate-github-org.shView on unpkg

scripts/hooks/kickoff-gate.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 1

High

Semantic Analysis Limited

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

scripts/hooks/kickoff-gate.js#virtual:normalized:round1View on unpkg

Findings

1 Critical4 High5 Medium5 Low

CriticalAi Agent Control Hijackscripts/install.js

HighInstall Time Lifecycle Scriptspackage.json

HighExternal Ai Agent Control Surface Mutationscripts/install.js

HighSemantic Analysis Limitedscripts/hooks/kickoff-gate.js#virtual:normalized:round1

HighPrevious Version Dangerous Deltascripts/install.js

MediumAmbiguous Install Lifecycle Scriptpackage.json

MediumNetwork

MediumEnvironment Vars

MediumShips Build Helperskills/migrate-github-org/migrate-github-org.sh

MediumStructural Risk Force Deep Review

LowScripts Present

LowFilesystem

LowHigh Entropy Strings

LowUrl Strings

LowNo License