The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates processing of untrusted external data, creating a surface for indirect prompt injection. * Ingestion points: Video transcripts generated from audio are ingested by the agent through the get-transcript tool. * Boundary markers: The instructions sent to the agent via sendToAgentChat do not implement delimiters or safety warnings to distinguish between system instructions and transcript content. * Capability inventory: The agent is empowered to perform state-changing actions such as update-recording, add-comment, and apply-edit based on the processed content. * Sanitization: There is no evidence of filtering or sanitizing the transcript text before it reaches the LLM.
[EXTERNAL_DOWNLOADS]: The skill utilizes external APIs for audio transcription. * Transcription fallback: If native or managed services are unavailable, the skill routes audio to Groq (api.groq.com), a well-known AI infrastructure service.
[CREDENTIALS_UNSAFE]: The skill mentions the use of API keys for authentication. * Evidence: Refers to BUILDER_PRIVATE_KEY and GROQ_API_KEY. * Handling: The skill provides instructions to manage these via a dedicated UI panel rather than hardcoding them, which follows security best practices.

ai-video-tools