Menu
Visual verification agent for ROM hacks. Compares emulator screenshots against "Golden Masters" to detect regressions, glitches, or UI errors.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | visual-verifier |
| description | Visual verification agent for ROM hacks. Compares emulator screenshots against "Golden Masters" to detect regressions, glitches, or UI errors. |
Save the current emulator frame as a reference.
capture title_screen: Saves assets/visual_masters/title_screen.png.Compare the current frame against a saved master.
verify title_screen: Reports similarity percentage (e.g., 99.5%).Analyze the frame for common SNES glitches.
capture room_0x1B (When you know the room is correct).verify room_0x1B.~/src/hobby/yaze/scripts/ai/visual_verifier.py.opencv-python, Pillow.Use the Agentic File System (AFS) context model for discovery, mounts, scratchpad memory, and file-based tool workflows. Use when a task involves .context operations, AFS CLI usage, or cross-agent context coordination.
Read and interpret the Link to the Past (ALTTP) disassembly with label/symbol context. Use when mapping ROM/CPU addresses to labels, searching usdasm/jpdasm sources, or cross-referencing routines with Hyrule Historian, z3dk label indexes, or z3ed output.
Train, evaluate, and maintain the scawful-echo persona and related avatar models (Echo/Memory/Muse). Use when working on persona voice, dataset prep, A/B testing, deployment, or tool-calling constraints for avatar models.
Autonomous navigation agent for Oracle of Secrets. Uses static map data and dynamic RAM state to localize Link and route him to destinations.
End-to-end Oracle of Secrets debugging using the Mesen2-OOS socket CLI, z3ed, and label/symbol tooling. Use for breakpoint/trace capture, memory inspection, savestate repro, overlays, or crash/regression triage (replaces mesen2-mcp and legacy Lua bridge workflows).
Expert guidance for LLM model training, synthetic data generation (SDE), Hierarchical MoE (H-MoE) management, and agentic evaluation. Use when setting up training runs (Vast.ai), generating datasets, or performing technical research PDF analysis for ROM hacking and general AI development.