@agent-link/agent@0.1.336
AgentLink local agent CLI
AI Security Review
scanned 14d ago · by lpm-firewall-aiSource inspection supports a clean verdict: the flagged agent paths are used for session and project data, and the postinstall migration is restricted to global installs and adopted versions. No concrete attack surface was established in the inspected files.
Decision evidence
public snapshot- The manifest routes the CLI through dist/cli.js and defines a postinstall hook; that hook runs its migration only when npm_config_global is true.
- The Codex code references the user’s ~/.codex/sessions directory and a package config file for session titles.
- The history code references the user’s ~/.claude/projects directory to parse session metadata.
- The memory code locates memory files beneath the user’s ~/.claude/projects directory, scoped to a project folder.
- Service artifact paths are selected for standard user service locations or an explicitly configured simulation root.
- The launcher checks that a resolved path stays inside its specified root before loading a pointer.
- The install bridge is gated on global npm installation; its migration requires an adopted version store and a newer stable version.
- The inspected agent configuration locations show session and project data paths; no credential forwarding or unrelated recipient was identified there.
Source & flagged code
6 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgPackage source references weak cryptographic algorithms.
dist/emergency-upgrade.jsView on unpkg · L4Findings
Affected versions and remediation
This report applies to @agent-link/agent@0.1.336.
See all versions of @agent-link/agent for other recorded verdicts.
- Review the evidence and your use of @agent-link/agent@0.1.336 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.