media-transcription
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use the
joelclawCLI tool to send events and monitor pipeline runs. These commands include shell logic for generating unique request identifiers. - [DATA_EXFILTRATION]: The skill accesses media files located on a specific NFS mount (
/Volumes/badass-media). This access is limited to the skill's primary function of transcribing media and does not involve accessing sensitive system credentials or files. - [EXTERNAL_DOWNLOADS]: The skill references an external code repository (
joelhooks/transcript-rig) as the underlying implementation. This repository is owned by the skill's author and is used to define the transcription logic. - [PROMPT_INJECTION]: The skill processes untrusted media files from external sources. While there is an inherent risk of indirect prompt injection if the resulting transcripts are used to influence further agent actions, the skill describes a decoupled architecture where transcription actors run as detached local processes with dedicated data validation steps.
Audit Metadata