Loading npm security reports…

sprag-cli@3.44.0: AI-agent control-surface risk | LPM Firewall

registry / sprag-cli / 3.44.0

sprag-cli@3.44.0

Sprag - a quality ratchet harness for AI coding agents: model-fitting delegation, ratchet rules, doc2md, Korean style gate. Same tool as claude-token-saver, under its new name.

AI Security Review

scanned 24d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. Installing the package automatically modifies the user's global Claude Code control surface. It installs package-controlled instructions and command hooks that run the package during later AI-agent events.

Static reason

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

Trigger

npm postinstall during package installation.

Impact

The package gains persistent execution and instruction influence across Claude Code sessions and projects.

Mechanism

Automatic global Claude Code skill, settings, hook, and instruction injection.

Policy narrative

The manifest invokes `sprag install` at postinstall and suppresses failures. That route writes a package-controlled skill and command hooks into the global Claude Code configuration, then appends a package-defined block to the global Claude Code instructions. The resulting hooks invoke the package at session start and user-prompt submission, creating persistent, unconsented influence over the AI-agent environment.

AI rationale

This is an automatic postinstall mutation of a broad, foreign AI-agent control surface, not an explicit user setup action. The persistent hooks and global instruction file establish concrete install-hook abuse.

Evidence

package.jsonbin/cli.jssrc/commands/install.jssrc/installer.js~/.claude/settings.json~/.claude/skills/claude-token-saver/SKILL.md~/.claude/CLAUDE.md

Decision evidence

Source & flagged code

17 flagged · loading source

package.jsonView file

•scripts.postinstall = node bin/cli.js install || true

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

•scripts.postinstall = node bin/cli.js install || true

Medium

Ambiguous Install Lifecycle Script

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

package.jsonView on unpkg

src/doc2md.cjsView file

28const crypto = require('node:crypto'); L29: const { spawn, spawnSync } = require('node:child_process'); L30: const ledger = require('./doc2md-ledger.cjs');

High

Child Process

Package source references child process execution.

src/doc2md.cjsView on unpkg · L28

•matchType = normalized_sha256 matchedPackage = claude-token-saver@3.43.0 matchedPath = src/doc2md.cjs matchedIdentity = npm:Y2xhdWRlLXRva2VuLXNhdmVy:3.43.0 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.

src/doc2md.cjsView on unpkg

24L25: const fs = require('node:fs'); L26: const os = require('node:os');

Medium

Dynamic Require

Package source references dynamic require/import behavior.

src/doc2md.cjsView on unpkg · L24

src/commands/upgrade.jsView file

20// resolution to find the same npm that installed us. L21: const child = spawn(cmd, { shell: true, stdio: 'inherit' }); L22: child.on('close', (code) => resolve(code === null ? 1 : code));

High

Shell

Package source references shell execution.

src/commands/upgrade.jsView on unpkg · L20

src/seed-rules.jsView file

53try { L54: const data = JSON.parse(readFileSync(seedStatePath(), 'utf8')); L55: return data && typeof data.decided === 'object' && data.decided ? data : { decided: {} };

Low

Weak Crypto

Package source references weak cryptographic algorithms.

src/seed-rules.jsView on unpkg · L53

src/commands/feedback.jsView file

56if (auth.status !== 0) return null; L57: const res = spawnSync('gh', ['issue', 'create', '-R', REPO, '--title', title, '--body', body], L58: { encoding: 'utf8', timeout: 30_000 }); L59: if (res.status !== 0) { debug('feedback:gh', res.stderr); return null; } L60: const url = String(res.stdout).trim().split('\n').pop(); ... L67: try { L68: const params = new URLSearchParams(); L69: params.set(FORM.message, `${title}\n\n${body}`); L70: params.set(FORM.meta, meta); L71: const res = await fetch(`https://docs.google.com/forms/d/e/${FORM.id}/formResponse`, { L72: method: 'POST',

High

Command Output Exfiltration

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

src/commands/feedback.jsView on unpkg · L56

21L22: import { spawnSync } from 'node:child_process'; L23: import { writeFileSync, mkdirSync } from 'node:fs'; ... L28: const REPO = 'rootstudioyaml/sprag'; L29: const ISSUES_URL = `https://github.com/${REPO}/issues`; L30: ... L47: `version: ${version}`, L48: `os: ${process.platform} ${os.release()}`, L49: `node: ${process.version}`, ... L58: { encoding: 'utf8', timeout: 30_000 }); L59: if (res.status !== 0) { debug('feedback:gh', res.stderr); return null; } L60: const url = String(res.stdout).trim().split('\n').pop();

High

Sandbox Evasion Gated Capability

Source gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.

src/commands/feedback.jsView on unpkg · L21

56Trigger-reachable command-output exfiltration chain: scripts.postinstall -> bin/cli.js -> src/commands/feedback.js L56: if (auth.status !== 0) return null; L57: const res = spawnSync('gh', ['issue', 'create', '-R', REPO, '--title', title, '--body', body], L58: { encoding: 'utf8', timeout: 30_000 }); L59: if (res.status !== 0) { debug('feedback:gh', res.stderr); return null; } L60: const url = String(res.stdout).trim().split('\n').pop(); ... L67: try { L68: const params = new URLSearchParams(); L69: params.set(FORM.message, `${title}\n\n${body}`); L70: params.set(FORM.meta, meta); L71: const res = await fetch(`https://docs.google.com/forms/d/e/${FORM.id}/formResponse`, { L72: method: 'POST',

bin/cli.jsView file

10Install-time code directly mutates a foreign AI-agent control surface: L104: sprag uninstall remove everything install added L105: sprag harness check score the harness setup in CLAUDE.md L106: sprag harness analyze run the harness transcript analysis manually ... L251: L252: // Subcommand: harness — manage the project's CLAUDE.md harness rules. L253: // sprag harness init # write CLAUDE.md (5 sections) + ratchet.md L254: // sprag harness uninit # remove harness block from CLAUDE.md (backup kept) L255: // sprag harness check # show 🅷 N/5 + which sections are missing ... L266: // sprag compact-window # status L267: // sprag compact-window set --global # pin 200k (~/.claude/settings.json) L268: // sprag compact-window set --p

Critical

Ai Agent Control Hijack

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

bin/cli.jsView on unpkg · L10

examples/statusline-command.ps1View file

•path = examples/statusline-command.ps1 kind = build_helper sizeBytes = 1619 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

examples/statusline-command.ps1View on unpkg

src/advice.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 4

High

Semantic Analysis Limited

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

src/advice.js#virtual:normalized:round1View on unpkg

src/korean-style.jsView file

•matchType = normalized_sha256 matchedPackage = claude-token-saver@3.43.0 matchedPath = src/korean-style.js matchedIdentity = npm:Y2xhdWRlLXRva2VuLXNhdmVy:3.43.0 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.

src/korean-style.jsView on unpkg

src/litellm-budget.jsView file

•matchType = normalized_sha256 matchedPackage = claude-token-saver@3.43.0 matchedPath = src/litellm-budget.js matchedIdentity = npm:Y2xhdWRlLXRva2VuLXNhdmVy:3.43.0 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.

src/litellm-budget.jsView on unpkg

src/model-alias.jsView file

•matchType = normalized_sha256 matchedPackage = claude-token-saver@3.43.0 matchedPath = src/model-alias.js matchedIdentity = npm:Y2xhdWRlLXRva2VuLXNhdmVy:3.43.0 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.

src/model-alias.jsView on unpkg

Source & flagged code

17 flagged

package.jsonView file

Lines 1-33json

1{

2 "name": "sprag-cli",

3 "version": "3.44.0",

4 "description": "Sprag - a quality ratchet harness for AI coding agents: model-fitting delegation, ratchet rules, doc2md, Korean style gate. Same tool as claude-token-saver, under its new name.",

5 "type": "module",

6 "bin": {

7 "sprag": "bin/cli.js",

8 "claude-token-saver": "bin/cli.js"

9 },

10 "scripts": {

11 "test": "node --test",

12 "docs:statusline": "node scripts/docs-statusline.mjs",

13 "postinstall": "node bin/cli.js install || true",

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg · L13

Medium

Ambiguous Install Lifecycle Script

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

package.jsonView on unpkg · L13

14 "publish:legacy": "node scripts/publish-legacy.mjs",

15 "sync:ko": "node scripts/sync-readme-ko.mjs",

16 "prepublishOnly": "node scripts/sync-readme-ko.mjs"

17 },

18 "files": [

19 "bin/",

20 "src/",

21 "presets/",

22 "examples/",

23 "README.md",

24 "README.ko.md",

25 "LICENSE"

26 ],

27 "engines": {

28 "node": ">=18"

29 },

30 "keywords": [

31 "sprag",

32 "ratchet",

33 "harness",

src/doc2md.cjsView file

Lines 4-48javascript

4 *

5 * A pptx or xlsx read straight into the context window is close to the worst

6 * thing a token-saving tool can allow: the bytes are unreadable to the model,

7 * so it either gets nothing useful or spends a fortune finding that out. This

8 * intercepts the Read, converts the file once, caches the result, and points

src/commands/upgrade.jsView file

Lines 1-40javascript

1/**

2 * Subcommand: upgrade — install the latest release with the package manager

3 * that put this copy on disk.

4 *

5 * sprag upgrade # refresh the check, then install

6 * sprag upgrade --print # show the command, run nothing

7 *

8 * This is the command the model runs after the user says yes to the

9 * session-start offer, so it prints the exact command it is about to execute

10 * before executing it: an install that writes outside the project should never

src/seed-rules.jsView file

Lines 33-73javascript

33 addModelRule,

34 composeRuleText,

35 loadModelRules,

36 modelRuleBaseText,

37 syncAllFiles,

38} from './model-rules.js';

39import {

40 findProjectRoot,

41 harnessListRules,

42 harnessPromote,

43

src/commands/feedback.jsView file

Lines 1-76javascript

1/**

2 * Subcommand: feedback — submit a bug report or feature request from the

3 * terminal (or from a Claude Code session) without opening a browser.

4 *

5 * sprag feedback "statusline이 IntelliJ에서 깨져요"

6 * sprag feedback --title "cache chip" "5m TTL 칩이 안 사라짐"

7 *

8 * Why this exists: GitHub issues require a logged-in browser session, and

9 * corporate networks often block github.com entirely. This command tries

10 * three transports in order and reports which one carried the message:

bin/cli.jsView file

10Install-time code directly mutates a foreign AI-agent control surface: L104: sprag uninstall remove everything install added L105: sprag harness check score the harness setup in CLAUDE.md L106: sprag harness analyze run the harness transcript analysis manually ... L251: L252: // Subcommand: harness — manage the project's CLAUDE.md harness rules. L253: // sprag harness init # write CLAUDE.md (5 sections) + ratchet.md L254: // sprag harness uninit # remove harness block from CLAUDE.md (backup kept) L255: // sprag harness check # show 🅷 N/5 + which sections are missing ... L266: // sprag compact-window # status L267: // sprag compact-window set --global # pin 200k (~/.claude/settings.json) L268: // sprag compact-window set --p

Critical

Ai Agent Control Hijack

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

bin/cli.jsView on unpkg · L10

examples/statusline-command.ps1View file

•path = examples/statusline-command.ps1 kind = build_helper sizeBytes = 1619 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

examples/statusline-command.ps1View on unpkg

src/advice.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 4

High

Semantic Analysis Limited

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

src/advice.js#virtual:normalized:round1View on unpkg

src/korean-style.jsView file

•matchType = normalized_sha256 matchedPackage = claude-token-saver@3.43.0 matchedPath = src/korean-style.js matchedIdentity = npm:Y2xhdWRlLXRva2VuLXNhdmVy:3.43.0 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.

src/korean-style.jsView on unpkg

src/litellm-budget.jsView file

•matchType = normalized_sha256 matchedPackage = claude-token-saver@3.43.0 matchedPath = src/litellm-budget.js matchedIdentity = npm:Y2xhdWRlLXRva2VuLXNhdmVy:3.43.0 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.

src/litellm-budget.jsView on unpkg

src/model-alias.jsView file

•matchType = normalized_sha256 matchedPackage = claude-token-saver@3.43.0 matchedPath = src/model-alias.js matchedIdentity = npm:Y2xhdWRlLXRva2VuLXNhdmVy:3.43.0 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.

src/model-alias.jsView on unpkg