@exulu/backend@4.3.3
**Exulu IMP** (Intelligence Management Platform) is an open-source TypeScript framework for building production-ready AI agent applications. It provides the backend infrastructure for managing agents, semantic search contexts, background job processing, a
AI Security Review
scanned 1mo ago · by lpm-firewall-aiSource establishes explicit user-command mutation of agent skill directories, which warrants a warning under the supplied agent policy. The inspected lifecycle behavior does not establish malicious agent hijacking.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
Downloading the bootstrap helper and explicitly invoking its install or update command with saved configuration.
Impact
Installed skills can influence agent behavior. Saved preferences guide subsequent explicit updates; no automatic update schedule was identified.
Mechanism
The helper downloads skill archives from the configured backend and copies or symlinks their contents into selected agent skill directories.
AI rationale
Source establishes explicit user-command mutation of agent skill directories, which warrants a warning under the supplied agent policy. The inspected lifecycle behavior does not establish malicious agent hijacking.
Evidence
dist/index.jsdist/index.cjs.agents/skills.claude/skills.windsurf/skills~/.config/imp/skills.json
Decision evidence
public snapshotAI called this Suspicious at 90.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- The downloadable helper installs remote skills into configured agent directories, using copies or symlinks.
- Both bundles list multiple agent skill destinations, including .agents/skills and .claude/skills.
- The backend serves the helper in a bootstrap ZIP rather than executing it during installation.
- The helper requires explicit install or update commands; saved configuration selects clients, scope and backend.
- The npm postinstall hook invokes a separate Python setup script and optional dependency cleanup.
Evidence against
- No automatic npm lifecycle mutation of agent skill directories was identified.
- The helper skips existing directories without its management marker, except symlinks.
- Authentication uses an API key supplied through the helper's own configuration; no harvested service session was identified.
- No covert exfiltration or malicious remote execution chain was established.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell
HighEntropyStringsUrlStrings
NoLicenseWildcardDependency
Source & flagged code
12 flaggedpackage.jsonView file
•scripts.preinstall = node -e "if (process.version !== 'v22.18.0') { console.error('❌ Wrong Node.js version. Expected v22.18.0, got ' + process.version); process.exit(1); }"
Critical
Red Install Lifecycle Script
Install-time lifecycle script matches a deterministic static-gate block pattern.
package.jsonView on unpkg•scripts.postinstall = node scripts/postinstall.cjs
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node scripts/postinstall.cjs
Medium
Ambiguous Install Lifecycle Script
Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgbin/backend.cjsView file
9L10: const { spawn } = require('child_process');
L11: const { existsSync } = require('fs');
High
bin/setup-python.cjsView file
•matchType = previous_version_dangerous_delta
matchedPackage = @exulu/backend@3.7.4
matchedIdentity = npm:QGV4dWx1L2JhY2tlbmQ:3.7.4
similarity = 0.684
summary = stored previous version shares package body but lacks this dangerous source file
Critical
Previous Version Dangerous Delta
This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
bin/setup-python.cjsView on unpkg13L14: const execAsync = promisify(exec);
L15:
High
dist/index.jsView file
•Manifest-reachable source captures an API credential, sends it to a fixed unofficial gateway, and persists that redirection in agent or shell configuration.
dist/index.js:
(req.headers["authorization"] || req.headers["x-api-key"]) ?? ""
const { baseUrl, authHeaders } = resolveLiteLLMTarget();
const endpoint = `${baseUrl}/v1/embeddings`;
// Optional attribution target for API keys (type "api"): tags requests
Some tools collect credentials (API keys, passwords, tokens) through a secure form shown directly to the user in the chat UI. Credentials are never entered in the conversation itse...
if (output && typeof output === "object" && output.oauth?.authorizationUrl) {
read them with the readFile tool. Files you produce yourself (via writeFile or via shell
this request (readFile, writeFile, bas
High
Credential Redirect Persistence
Manifest-reachable source sends a prompted API credential to a fixed unofficial gateway and persists the redirection.
dist/index.jsView on unpkg107var redisServer = {
L108: host: process.env.REDIS_HOST ?? "",
L109: port: process.env.REDIS_PORT ?? "",
...
L374: };
L375: var resolveLiteLLMConfigPath = () => process.env.LITELLM_CONFIG_PATH ?? resolve(process.cwd(), "./config.litellm.yaml");
L376: var parseEmbeddingModels = (configPath) => {
...
L532: contentType: type,
L533: metadata: {
L534: ...metadata,
...
L574: const text = this.encoder.decode(tokens);
L575: return new TextDecoder().decode(text);
L576: }
High
Command Output Exfiltration
Source combines command execution, command-output handling, and outbound requests; review data flow before blocking.
dist/index.jsView on unpkg · L107107var redisServer = {
L108: host: process.env.REDIS_HOST ?? "",
L109: port: process.env.REDIS_PORT ?? "",
...
L374: };
L375: var resolveLiteLLMConfigPath = () => process.env.LITELLM_CONFIG_PATH ?? resolve(process.cwd(), "./config.litellm.yaml");
L376: var parseEmbeddingModels = (configPath) => {
...
L532: contentType: type,
L533: metadata: {
L534: ...metadata,
...
L574: const text = this.encoder.decode(tokens);
L575: return new TextDecoder().decode(text);
L576: }
Low
Weak Crypto
Package source references weak cryptographic algorithms.
dist/index.jsView on unpkg · L107ee/python/setup.shView file
•path = ee/python/setup.sh
kind = build_helper
sizeBytes = 13359
magicHex = [redacted]
Medium
Ships Build Helper
Package ships non-JavaScript build or shell helper files.
ee/python/setup.shView on unpkgee/python/transcription/tests/test_server.pyView file
•path = ee/python/transcription/tests/test_server.py
kind = payload_in_excluded_dir
sizeBytes = 3264
magicHex = [redacted]
High
Payload In Excluded Dir
Package hides binary, compressed, or executable-looking payloads in test/fixture/hidden paths.
ee/python/transcription/tests/test_server.pyView on unpkgdist/chunk-D4PCOOL5.jsView file
•stage = ast_semantic_analysis; reason = ast_alias_growth_limit_exceeded; limitedFiles = 4
High
Semantic Analysis Limited
A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.
dist/chunk-D4PCOOL5.jsView on unpkgFindings
2 Critical7 High6 Medium7 Low
CriticalRed Install Lifecycle Scriptpackage.json
CriticalPrevious Version Dangerous Deltabin/setup-python.cjs
HighInstall Time Lifecycle Scriptspackage.json
HighChild Processbin/backend.cjs
HighShellbin/setup-python.cjs
HighCredential Redirect Persistencedist/index.js
HighCommand Output Exfiltrationdist/index.js
HighPayload In Excluded Diree/python/transcription/tests/test_server.py
HighSemantic Analysis Limiteddist/chunk-D4PCOOL5.js
MediumAmbiguous Install Lifecycle Scriptpackage.json
MediumNetwork
MediumEnvironment Vars
MediumShips Build Helperee/python/setup.sh
MediumStructural Risk Force Deep Review
MediumWildcard Dependency
LowNon Install Lifecycle Scripts
LowScripts Present
LowWeak Cryptodist/index.js
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
LowNo License
Affected versions and remediation
This report applies to @exulu/backend@4.3.3.
See all versions of @exulu/backend for other recorded verdicts.
- Review the evidence and your use of @exulu/backend@4.3.3 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: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.