halfcycle@0.3.36
Your coding agents are the engineers. Halfcycle is the rest of the team: product owner, architect, planner and independent reviewer for Claude Code.
AI Security Review
scanned 12d ago · by lpm-firewall-aiThe telemetry path sends a per-engagement credential to a service-provided URL without a host restriction or separate opt-in. Since the source does not show an unrelated recipient, this is an unresolved credential exposure risk rather than confirmed malicious behavior.
Decision evidence
public snapshot- The package exposes a CLI through `dist/bin.js`.
- CLI credential requests send bearer credentials to the configured service origin; the inspected callers cover engagement creation and joining, build records, and phase operations.
- The CLI accepts a telemetry URL from the service response and stores it for the guard runner without restricting its host.
- The runner conditionally posts evaluation data and a per-engagement bearer token to that URL; no recipient-specific opt-in or destination allowlist is shown.
- Telemetry is sent only when the outcome is configured and both the URL and token are present; the runner also omits the token when the credential is refused.
- The inspected source does not establish that the service-provided telemetry recipient is unrelated to Halfcycle, so theft or malicious intent is not confirmed.
Source & flagged code
5 flaggedPackage source references child process execution.
bin/bin.bundle.mjsView on unpkg · L15662This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
dist/bin.jsView on unpkgFindings
Affected versions and remediation
This report applies to halfcycle@0.3.36.
See all versions of halfcycle for other recorded verdicts.
- Review the evidence and your use of halfcycle@0.3.36 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.