tokenusage-cli@0.2.1
Self-hosted AI coding token usage collector for token-usage (forked from TokenTracker, MIT).
AI Security Review
scanned 6d ago · by lpm-firewall-aiThe flagged persistence and configuration changes are tied to the explicit init workflow, while the credential flow inspected sends Claude credentials to Anthropic for usage data. The manifest has no install hook, and the reported agent-configuration pointer is a local cache write.
Decision evidence
public snapshot- The manifest exposes a CLI and sets only a prepublishOnly test script; it declares no install, preinstall, or postinstall hook.
- The init command writes tracker configuration and sets up integrations as part of an explicit user command.
- On Windows, init can register an hourly sync task; the source documents the --no-task option and pause/resume controls.
- Claude usage code reads a Claude OAuth token and sends it as a bearer token to api.anthropic.com, the credential's own service.
- The categorizer writes a bounded local cache under ~/.tokenusage/cache/claude-categorizer/.
- The OpenClaw hook setup is a callable integration routine; the inspected CLI startup does not invoke it automatically.
- The scanner’s agent-configuration pointer resolves to a local categorizer cache write, not an AI-agent settings mutation.
Source & flagged code
5 flaggedPackage source references weak cryptographic algorithms.
src/lib/claude-categorizer.jsView on unpkg · L335A manifest entrypoint or package-local install chain reaches persistence behavior.
Findings
Affected versions and remediation
This report applies to tokenusage-cli@0.2.1.
See all versions of tokenusage-cli for other recorded verdicts.
- Review the evidence and your use of tokenusage-cli@0.2.1 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.