The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies a potential surface for indirect prompt injection in references/session-orchestration.md. The DEBRIEF_PROMPT interpolates raw interview transcripts into a system-level prompt for the Claude AI model. This allows data originating from user speech to potentially influence the automated evaluation logic.
Ingestion points: references/session-orchestration.md via the DebriefInput interface.
Boundary markers: The prompt uses markdown headers (e.g., ## Transcript) but lacks explicit instructions to the model to disregard instructions embedded within the transcript.
Capability inventory: The orchestrator performs network calls to the Anthropic API to generate debriefs.
Sanitization: No specific filtering or sanitization of the transcript content is implemented before interpolation.
[EXTERNAL_DOWNLOADS]: The proctoring engine setup in references/proctor-engine-setup.md fetches MediaPipe Face Mesh scripts and models from the cdn.jsdelivr.net content delivery network. This is a standard and expected method for integrating well-known machine learning libraries into a web-based application.
[COMMAND_EXECUTION]: The setup instructions in SKILL.md involve standard shell commands such as git clone, npm install, and npx supabase. These commands are necessary for initializing the platform's infrastructure and are consistent with the skill's primary function as a development guide.

interview-simulator