dsh-watch-ears@0.1.0
Hearing for DSH: capture system audio and the microphone, transcribe both with a resident listener so a text-only model can listen continuously.
AI Security Review
scanned 4d ago · by lpm-firewall-aiThe package automatically reuses a key from another provider's configuration and can send it to BigModel when the provider base URL is missing. This is credential exfiltration even though audio transcription itself matches the package's purpose.
Decision evidence
public snapshot- The package reads an API key from the active provider in ~/.modlens/config.json, defaulting to the openai provider.
- When that provider has no baseUrl, the key is paired with https://open.bigmodel.cn/api/paas/v4.
- The default ASR backend is auto and selects cloud when local transcription is unavailable.
- Cloud transcription sends the configured API key as a Bearer credential to the configured baseUrl.
- The cloud transcription request also uploads the recorded WAV audio.
- A hear request records audio and passes it to the transcription function.
- The package is described and implemented as an audio transcription tool, so sending audio to an ASR service is part of its purpose.
- Listening is initiated through an explicit hear or listen command/tool call; no install lifecycle scripts are declared.
Source & flagged code
7 flaggedSource combines command execution, command-output handling, and outbound requests; review data flow before blocking.
bin/companion.mjsView on unpkg · L4Findings
Affected versions and remediation
This report applies to dsh-watch-ears@0.1.0.
See all versions of dsh-watch-ears for other recorded verdicts.
- Avoid installing dsh-watch-ears@0.1.0. Remove it from direct dependencies and check your lockfile for transitive copies.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
- If this version ran, investigate the affected machine and build environment. Rotate credentials it could access and rebuild from a trusted environment.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.