@phnx-labs/agents-cli@1.22.68
One CLI for all your AI coding agents - versions, config, cloud dispatch, sessions, and teams (now with first-class Grok Build CLI support)
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThis is a first-party agent extension lifecycle risk with unprompted persistence and user-home mutation. The source does not show the concrete malicious chains required for a publish block.
Decision evidence
public snapshot- The manifest runs a postinstall script automatically.
- The lifecycle script creates a user-level .agents directory even for local installs.
- Global installation can add its executable directory to a login shell profile without an interactive consent prompt.
- The lifecycle script starts an always-on daemon; on Linux it writes, enables, and starts a user systemd service.
- On macOS, installation replaces a user-scoped application bundle with the bundled helper.
- No runtime dependency on @phnx-labs/agents-cli itself appears in package.json, so no nested self-lifecycle chain is present.
- The inspected lifecycle code contains no network request or credential exfiltration path.
- Agent configuration cloning and network features are attached to explicit CLI commands rather than the postinstall path.
Source & flagged code
23 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgSource file is highly similar to a previously finalized malicious package; route for source-aware review.
Findings
Affected versions and remediation
This report applies to @phnx-labs/agents-cli@1.22.68.
See all versions of @phnx-labs/agents-cli for other recorded verdicts.
- Review the evidence and your use of @phnx-labs/agents-cli@1.22.68 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.