@tonbo/yield@0.3.9
Tonbo Yield: capture your agents' sessions on every machine and list them from any.
AI Security Review
scanned 2d ago · by lpm-firewall-aiThe apparent OpenAI and Anthropic credential indicators are test-fixture strings, while the bundled authentication uses Tonbo credentials and its own API. Session capture and runtime metrics are described in the README as core functionality, with metrics content limits and an opt-out.
Decision evidence
public snapshot- The package describes its purpose as capturing Claude Code and Codex sessions and listing them across the user's devices.
- The bundled credential loader uses a Tonbo-specific key or saved Tonbo login; the default API host is api.tonbo.dev.
- The OpenAI credential pattern appears inside a test fixture showing an example line and its filtered output, not credential collection logic.
- README documents the runtime metrics as content-free and says they exclude tasks, prompts, paths, and reasons.
- The CLI captures agent sessions and snapshots Git work trees, which is substantial data handling but matches the stated package purpose.
- The package records runtime metrics by default; README describes the metrics and provides TONBO_YIELD_METRICS=off to disable them.
- The bundle contains a curl-to-shell installer command for upgrading, but it targets the package's own yield.tonbo.io host.
Source & flagged code
5 flaggedPackage contains a critical-looking secret pattern.
dist/bin/yield.jsView on unpkg · L3561GitHub personal access token in dist/bin/yield.js
dist/bin/yield.jsView on unpkg · L3561Manifest-reachable source sends a prompted API credential to a fixed unofficial gateway and persists the redirection.
Findings
Affected versions and remediation
This report applies to @tonbo/yield@0.3.9.
See all versions of @tonbo/yield for other recorded verdicts.
- Review the evidence and your use of @tonbo/yield@0.3.9 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.