speech-to-text
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DATA_EXPOSURE]: The skill transmits data to an external API endpoint (
eachsense-agent.core.eachlabs.run) for transcription processing. - The communication is necessary for the skill's primary function of speech-to-text conversion.
- It correctly uses placeholders like
$EACHLABS_API_KEYandYOUR_EACHLABS_API_KEY, encouraging users to manage secrets through environment variables rather than hardcoding them. - [INDIRECT_PROMPT_INJECTION]: The skill processes audio and video content from external URLs, which creates a surface for instructions embedded in media to influence agent behavior.
- Ingestion points: Media URLs are provided via the
image_urlparameter in JSON payloads (e.g., inSKILL.md). - Boundary markers: The skill uses structured JSON payloads to separate system/user instructions from the data resources (URLs).
- Capability inventory: The skill uses
Bash(curl)andWebFetchto transmit data and fetch resources. - Sanitization: There is no explicit sanitization described for the content returned by the transcription service before it is processed by the agent.
- [COMMAND_EXECUTION]: The skill documentation includes examples using
Bash(curl)to interact with the service. - These commands are standard for API interaction and do not involve administrative privileges or dangerous flag combinations.
Audit Metadata