The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for building conversational agents that ingest untrusted user speech via speech-to-text (STT) services, which serves as a potential vector for indirect prompt injection.
Ingestion points: User audio is captured and transcribed using sarvam.STT or SarvamSTTService (SKILL.md).
Boundary markers: The provided code templates do not implement specific delimiters or "ignore instructions" system prompts to encapsulate the transcribed user text.
Capability inventory: The agents are designed to process transcribed text through an LLM (e.g., sarvam-105b) and generate voice responses via TTS.
Sanitization: The examples do not demonstrate explicit sanitization or filtering of the transcribed text before it is processed by the language model.
[EXTERNAL_DOWNLOADS]: The skill references documentation and integration guides hosted on the developer's official domain.
Evidence: Links to docs.sarvam.ai for framework integration guides and rate limit information (SKILL.md).
[SAFE]: The skill encourages secure development practices by recommending the use of environment variables (SARVAM_API_KEY) for sensitive credentials rather than hardcoding them.

voice-agents