groq-inference
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources including user prompts, audio files, and images, which presents an indirect prompt injection surface.\n
- Ingestion points: Operations such as
transcribe,analyze_image, andchatinSKILL.mdandtemplates/groq-client.pyingest data that could contain hidden instructions (e.g., in audio or image metadata/content).\n - Boundary markers: The skill lacks explicit delimiters or instructions to the LLM to ignore potentially malicious instructions embedded within the data being processed.\n
- Capability inventory: The skill has access to network resources (Groq API), file system reads (for media processing), and local function execution via the tool-calling pattern.\n
- Sanitization: There is no evidence of sanitization or filtering of data returned from transcription or vision tasks before it is used in further logic.\n- [EXTERNAL_DOWNLOADS]: The documentation suggests the use of several external Python libraries for extended functionality.\n
- Evidence: References to
groq,tenacity,deepgram-sdk,assemblyai,cartesia,elevenlabs, andanthropicare found inSKILL.mdandreference/audio-speech.md.\n- [COMMAND_EXECUTION]: The skill instructs the agent to perform a file system write operation to record execution metrics.\n - Evidence:
SKILL.mdcontains an 'Emit Outcome Sidecar' section directing the agent to write a JSON file to~/.claude/skill-analytics/last-outcome-groq-inference.jsonupon completion.
Audit Metadata