registry / gm-plugkit / 2.0.2423
gm-plugkit@2.0.2423
Bootstrap and daemon-spawn tool for gm plugkit binary. Downloads the correct platform wasm, verifies SHA256, and launches agentplug-runner (the native wasm host) as the spool watcher daemon.
AI Security Review
scanned 3mo ago · by lpm-firewall-aiThe source establishes a real, user-invoked global AI-agent extension and remote payload update path. With no install hook or observed exfiltration, policy calls for a warning rather than a publish block.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
User runs gm-plugkit or calls ensureReady.
Impact
A remote or bundled instruction payload can influence compatible agents across projects; the daemon executes externally supplied WASM via a native runner.
Mechanism
Remote agent-skill provisioning plus downloaded WASM daemon launch.
AI rationale
The source establishes a real, user-invoked global AI-agent extension and remote payload update path. With no install hook or observed exfiltration, policy calls for a warning rather than a publish block.
Evidence
package.jsonbootstrap.jsbootstrap-shared.jscli.jsSKILL.md$HOME/.agents/skills/<skill>/SKILL.md$HOME/.claude/skills/<skill>/SKILL.md<project>/CLAUDE.md<project>/AGENTS.md<project>/.gm/next-step.md$HOME/.gm-tools/plugkit.wasm
Network endpoints5
api.github.com/repos/AnEntrypoint/gm/contents/skills?ref=mainraw.githubusercontent.com/AnEntrypoint/gm/main/skills/<skill>/SKILL.mdapi.github.com/repos/AnEntrypoint/plugkit-bin/releases?per_page=50registry.npmjs.org/gm-plugkit/latestgithub.com/AnEntrypoint/plugkit-bin/releases/download/v<version>
Decision evidence
public snapshotAI called this Suspicious at 97.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- bootstrap.js installs skills into $HOME/.agents/skills and $HOME/.claude/skills.
- bootstrap-shared.js adds @AGENTS.md to CLAUDE.md and @.gm/next-step.md to project AGENTS.md.
- bootstrap.js fetches remotely discovered SKILL.md files and writes them to agent skill directories without a pinned content hash.
- SKILL.md instructs agents to avoid ending turns, assume standing approval, and follow a remote instruction stream.
- bootstrap.js downloads a WASM payload, copies it to $HOME/.gm-tools, and cli.js starts agentplug-runner.
Evidence against
- package.json has no preinstall, install, or postinstall lifecycle hook.
- The control-surface changes occur when the user invokes the CLI/ensureReady, not on npm installation.
- No credential harvesting or direct data-exfiltration code was found.
- Declared network hosts and payload purpose are package-aligned.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetwork
Source & flagged code
3 flagged · loading sourcebootstrap-shared.jsView file
5const os = require('os');
L6: const { spawnSync } = require('child_process');
L7: const { pidAlive, sha256OfFileSync } = require('./gm-process');
High
Child Process
Package source references child process execution.
bootstrap-shared.jsView on unpkg · L5bootstrap.jsView file
7const crypto = require('crypto');
L8: const { spawn, spawnSync } = require('child_process');
L9: const { sha256OfFile, sha256OfFileSync } = require('./gm-process');
...
L37:
L38: const wrapperDir = __dirname;
L39:
...
L44: try {
L45: const ownPkg = JSON.parse(fs.readFileSync(path.join(__dirname, 'package.json'), 'utf-8'));
L46: if (ownPkg && ownPkg.version) {
...
L167: const dir = gmToolsDir();
L168: const names = process.platform === 'win32'
L169: ? ['agentplug-runner.exe']
Critical
Download Execute
Source downloads or fetches remote code and executes it.
bootstrap.jsView on unpkg · L77Trigger-reachable chain: manifest.main -> index.js -> bootstrap.js
L7: const crypto = require('crypto');
L8: const { spawn, spawnSync } = require('child_process');
L9: const { sha256OfFile, sha256OfFileSync } = require('./gm-process');
...
L37:
L38: const wrapperDir = __dirname;
L39:
...
L44: try {
L45: const ownPkg = JSON.parse(fs.readFileSync(path.join(__dirname, 'package.json'), 'utf-8'));
L46: if (ownPkg && ownPkg.version) {
...
L167: const dir = gmToolsDir();
L168: const names = process.platform === 'win32'
L169: ? ['agentplug-runner.exe']
Findings
2 Critical1 High3 Medium3 Low
CriticalDownload Executebootstrap.js
CriticalTrigger Reachable Dangerous Capabilitybootstrap.js
HighChild Processbootstrap-shared.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to gm-plugkit@2.0.2423.
See all versions of gm-plugkit for other recorded verdicts.
- Review the evidence and your use of gm-plugkit@2.0.2423 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.