cli
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides commands to download and execute official installation scripts from the vendor's domain (e.g.,
curl -fsSL https://deepgram.com/install.sh | shandiwr https://deepgram.com/install.ps1 -useb | iex). These are recognized as the primary vendor's distribution channels. - [COMMAND_EXECUTION]: The skill instructs the agent on how to use the
dg,deepctl, anddeepgrambinaries to perform operations such as transcription, speech synthesis, and project management. - [CREDENTIALS_UNSAFE]: The documentation warns users that API keys may be stored in cleartext in
config.yamlif a secure OS keyring is not available. It advises treating this file as a secret and using ephemeral environments in CI/CD pipelines. - [INDIRECT_PROMPT_INJECTION]: As the skill ingests external data for transcription (e.g.,
dg listen https://dpgr.am/spacewalk.wav) and text analysis (dg read --file feedback.txt), it possesses a surface for indirect prompt injection where malicious instructions could be embedded in the audio or text content. However, this is a standard risk for processing tools and is mitigated by the skill's specific operational context.
Audit Metadata