pincushion-mcp@1.11.26
Visual feedback as agent work packets. Stakeholders drop pins on your live app; your AI coding agent reads each pin (selector, screenshot, DOM, thread, acceptance criteria) via MCP and ships the fix.
AI Security Review
scanned 1mo ago · by lpm-firewall-aiThe package contains a concrete agent-facing instruction to delete files without confirmation. Its lack of lifecycle hooks and project-local setup behavior do not remove that prompt-level risk.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
A user or agent invokes the implement prompt.
Impact
An agent may delete repository lock files without checking whether they are legitimate active locks.
Mechanism
The prompt labels a Git lock-file deletion command mandatory and says to run it without permission.
AI rationale
The package contains a concrete agent-facing instruction to delete files without confirmation. Its lack of lifecycle hooks and project-local setup behavior do not remove that prompt-level risk.
Evidence
server.js.git/HEAD.lock.git/index.lock.git/refs/heads/*.lock.git/config.lock
Decision evidence
public snapshotAI called this Suspicious at 91.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- The package exposes server.js as its executable entrypoint.
- The implement prompt directs an agent to delete Git lock files immediately and without permission.
Evidence against
- No npm preinstall, install, or postinstall hook is declared.
- The editor setup helper is limited to an explicitly selected project-local configuration and avoids global configuration.
- Telemetry hashes the license-key identifier before capture.
Behavioral surface
ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNetworkShell
HighEntropyStringsTelemetryUrlStrings
NoLicense
Source & flagged code
10 flaggedmobile-wall.mjsView file
17// injectable so both modes are testable without a simulator.
L18: import { execFileSync } from 'node:child_process';
L19: import {
High
mobile-wall.mjs#virtual:normalized:round1View file
121function publicUrlFrom(text) {
L122: const m = /(https?:\/\/[^\s"'<>]+)/.exec(String(text || ''));\n return m ? m[1].replace(/[),.;]+$/, '') : '';\n}\n\nasync function fetchCloudAnnotations(ctx, { baseUrl, key, proje...
L123: }
Critical
Same File Env Network Execution
A single source file combines environment access, network access, and code or shell execution with blocking evidence.
mobile-wall.mjs#virtual:normalized:round1View on unpkg · L12117// injectable so both modes are testable without a simulator.
L18: import { execFileSync } from "node:child_process";
L19: import {
...
L32: const DEFAULT_OUT = "crit-wall";
L33: const DEFAULT_FUNCTION_BASE = "https://dpsqzszdviltqvethxbr.supabase.co/functions/v1";
L34:
...
L78: function isSignedIn() {
L79: if (process.env.PINCUSHION_LICENSE_KEY) return true;
L80: try { return !!readFileSync(join(homedir(), ".pincushion", "license-key"), "utf8").trim(); } catch { return false; }
L81: }
...
L97: const firstLine = (e) => String(e.message || e).split("\n")[0];
L98: const execErrText = (e) => [e?.stderr, e?.stdout, e?.message].filter(Boolean).join("\n");
Findings
4 Critical7 High4 Medium6 Low
CriticalSame File Env Network Executionmobile-wall.mjs#virtual:normalized:round1
CriticalCredential Exfiltrationserver.js
CriticalDownload Executemobile-wall.mjs#virtual:normalized:round1
CriticalTrigger Reachable Dangerous Capabilityserver.js
HighChild Processmobile-wall.mjs
HighShell
HighCommand Output Exfiltrationmobile-wall.mjs#virtual:normalized:round1
HighCopied Package Dependency Bridgelib/posthog.js
HighTrigger Reachable Credential Exfiltrationserver.js
HighTrigger Reachable Command Output Exfiltrationmobile-wall.mjs#virtual:normalized:round1
HighSemantic Analysis Limitedlib/license.js#virtual:normalized:round1
Affected versions and remediation
This report applies to pincushion-mcp@1.11.26.
See all versions of pincushion-mcp for other recorded verdicts.
- Review the evidence and your use of pincushion-mcp@1.11.26 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.