linkshell-cli@0.5.32
Remote terminal bridge for Claude Code / Codex / Gemini / Copilot — control from your phone
AI Security Review
scanned 4mo ago · by lpm-firewall-aiStatic inspection found high-risk primitives, but they are runtime/user-invoked and consistent with the documented remote terminal and agent workspace product. The install hook is limited to local node-pty permission repair and no credential harvesting, covert exfiltration, persistence, or lifecycle AI-agent control mutation was confirmed.
Decision evidence
public snapshot- src/runtime/bridge-session.ts writes Claude/Codex/Gemini/Copilot hook config during linkshell start
- src/runtime/acp/claude-stream-json-client.ts can run Claude fallback with bypassPermissions for remote prompts
- src/commands/upgrade.ts runs npm install -g only when user invokes linkshell upgrade
- scripts/postinstall.mjs only chmods node-pty spawn-helper candidates; no network or payload fetch
- README.md describes remote terminal/agent bridge, screen sharing, login, gateway, and hook-like agent workspace behavior
- Network calls are package-aligned: Supabase auth, itool.tech login, npm registry update check, configured gateway WebSocket
- Auth token storage is local ~/.linkshell/auth.json with mode 0600 and used for LinkShell gateway/auth APIs
- Shell/PTY spawning is the core advertised CLI bridge behavior and is activated by explicit runtime commands
Source & flagged code
15 flaggedPackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgPackage contains a critical-looking secret pattern.
dist/cli/src/auth.jsView on unpkg · L39Findings
Affected versions and remediation
This report applies to linkshell-cli@0.5.32.
See all versions of linkshell-cli for other recorded verdicts.
- Review the evidence and your use of linkshell-cli@0.5.32 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.