@zaip-ai/cli@1.0.8
CLI oficial para instalar e configurar o ZAIP-AIOS em qualquer repositório
AI Security Review
scanned 3mo ago · by lpm-firewall-aiNot malicious by source evidence: no automatic install trigger, credential theft, or exfiltration was confirmed. Warn because the explicit installer configures high-privilege agent control surfaces and hooks.
Static reason
No blocking static signals were detected.
Trigger
User runs `zaip init` without `--project-only`.
Impact
Can alter Codex agent permissions and hook behavior for the initialized project; global Codex configuration receives a hook entry.
Mechanism
Installs agent templates, privileged Codex settings, and command hooks.
AI rationale
Not malicious by source evidence: no automatic install trigger, credential theft, or exfiltration was confirmed. Warn because the explicit installer configures high-privilege agent control surfaces and hooks.
Evidence
package.jsondist/commands/init.jsdist/services/install-agents-template.jsdist/services/adapters.jsdist/services/global-codex-config.jstemplates/.agents/design/packages/windows-launcher/Program.cs.agents/**.codex/config.toml.codex/hooks/zaip-pre-tool-use.mjs.codex/hooks/zaip-subagent-lifecycle.mjs
Network endpoints1
developers.openai.com/mcp
Decision evidence
public snapshotAI called this Suspicious at 90.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- `zaip init` writes templates and then reconciles global Codex configuration by default.
- Generated project Codex config grants a default profile `:root` write access and unrestricted network.
- The generated config installs PreToolUse and Subagent lifecycle command hooks.
- Bundled Windows executables are template launchers; source shows the design launcher starts a workspace-local Node script.
Evidence against
- No preinstall, install, or postinstall lifecycle hook is declared.
- CLI execution requires an explicit `zaip init` command.
- Template files are hash-verified before installation.
- No credential harvesting or package-initiated exfiltration was confirmed.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemShell
HighEntropyStringsUrlStrings
NoLicense
Source & flagged code
6 flaggedtemplates/.agents/design/zaip-design.exeView file
•path = templates/.agents/design/zaip-design.exe
kind = payload_in_excluded_dir
sizeBytes = 7680
magicHex = [redacted]
High
Payload In Excluded Dir
Package hides binary, compressed, or executable-looking payloads in test/fixture/hidden paths.
templates/.agents/design/zaip-design.exeView on unpkg•path = templates/.agents/design/zaip-design.exe
kind = native_binary
sizeBytes = 7680
magicHex = [redacted]
Medium
Ships Native Binary
Package ships native binary artifacts.
templates/.agents/design/zaip-design.exeView on unpkgtemplates/.agents/design/packages/windows-launcher/build-design-launcher.ps1View file
•path = templates/.agents/design/packages/windows-launcher/build-design-launcher.ps1
kind = build_helper
sizeBytes = 3452
magicHex = [redacted]
Findings
1 Critical1 High6 Medium5 Low
CriticalTrigger Reachable Excluded Directory Payloaddist/services/install-agents-template.js
HighPayload In Excluded Dirtemplates/.agents/design/zaip-design.exe
MediumEnvironment Vars
MediumShips Native Binarytemplates/.agents/design/zaip-design.exe
MediumShips Build Helpertemplates/.agents/design/packages/windows-launcher/build-design-launcher.ps1
MediumOversized Source Filetemplates/.agents/runtime/bin/zaip.mjs
MediumOversized Cli Entrypointtemplates/.agents/runtime/bin/zaip.mjs
MediumStructural Risk Force Deep Review
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to @zaip-ai/cli@1.0.8.
See all versions of @zaip-ai/cli for other recorded verdicts.
- Review the evidence and your use of @zaip-ai/cli@1.0.8 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.