ghchinoy

Diagnose and repair beads (bd) issue-tracker problems caused by its Dolt backend — including engine-mode mismatches (embedded vs server), database name incompatibilities, DATABASE MISMATCH repo-ID errors, the "auto-backup failed / table file not found" corruption that silently reverts writes, lock contention from multiple `dolt sql-server` processes (across projects or host/container), the hook-timeout stash-wipe that destroys untracked files on commit, orphaned dolt sql-server process leaks, and schema version skew in BOTH directions — a newer bd client refusing to auto-apply migrations to a remote-backed database, AND an older client stranded behind a database another agent/machine already migrated forward (bd doctor reports "database is at vX, binary knows up to vY (N migrations ahead)" and writes fail with "Field 'id' doesn't have a default value"). Use when bd won't start, a daemon-error file is present, bd updates don't persist, issues.jsonl is out of sync with bd show output, untracked files vanish aft

Create, assess, and improve README.md files for public GitHub projects. Scores an existing README against a weighted rubric, gives targeted section-by-section fixes, and can check strict standard-readme spec compliance. Use when the user asks to write a README, review/grade a README, improve an existing README, or check standard-readme compliance.

Edit technical blog posts to remove AI writing patterns and preserve human voice. Use when drafting or reviewing developer-facing posts, migration guides, and engineering write-ups.

Sub-skill for composing high-fidelity music tracks via Google DeepMind Lyria and verifying container formats.

Sub-skill for validating the artist.json metadata contract, copying files, compiling typescript/Vite templates, and verifying static client builds.

Sub-skill for drafting virtual artist backstories, mini-albums, theme-based color palettes, track metadata, and lyrics.

Primary orchestrator skill for creating concept-driven AI Pop Artists. Use when a user asks to generate a new artist profile, album narrative, audio-visual assets, and local web-player dashboard.

Sub-skill for generating cohesive album cover art and track illustrations using NanoBanana.

Guides agents on preparing datasets, running parameter-efficient LoRA fine-tuning, and tracking telemetry metrics using the mlxtune CLI on Apple Silicon. Use when an agent wants to fine-tune standard Gemma 4 base models on text instructions.

Guides agents on exporting fine-tuned Gemma models to multi-format mobile runtimes, comparing memory and hardware target characteristics, and implementing dynamic sandbox download workflows.

Guides agents on preparing multimodal datasets in interleaved ChatML format, executing audio/vision fine-tuning with mlx-vlm, targeting specific layers, and performing key-sanitized weights fusion to avoid silent blank weights loading.

Guides agents on fine-tuning Gemma 4 Quantization-Aware Training (QAT) unquantized checkpoints, maintaining mathematical alignment during LoRA, and exporting to GGUF using strict q4_0 parameters to avoid post-training quantization drift on edge devices.

Use when creating new Elvish words. Provides a structured choice between "functional/practical" neologisms and "poetic/metaphorical" neologisms, guided by precise, diachronic phonological sound laws.

Generates grammatically correct, historically authentic Tolkien Elvish names for people, places, weapons, or stars. Use when creating fictional names, compounding roots, or translating modern names into Elvish. Invokes eldamo-remote tools to find matching vocabulary roots and grammatical suffixes.

Translates English phrases into Tolkien's languages (Quenya, Sindarin, Adûnaic). Use when translating text, composing Elvish poetry, or validating grammar. Invokes eldamo-remote tools to find dictionary words, historical roots, and morphological details.

Synthesize high-fidelity English speech and perform zero-shot voice cloning locally on macOS GPU via MLX. Use when a user requests text-to-speech, speaker voice cloning with an audio reference, custom speaker indexing, adjusting sampling parameters, or bypassing watermarks.

Evaluate, transcribe, and mathematically verify synthesized audio and speech outputs. Use when measuring Log-Mel Spectrogram MAE, Temporal Envelope Correlation, or when running Vertex AI Gemini or offline Gemma 4 multimodal quality evaluations.

Port, migrate, and validate PyTorch neural network modules to Apple Silicon MLX format. Use when translating attention layers, customizing Llama RoPE scale factors, setting up step-by-step autoregressive KV caches, or mapping PyTorch checkpoint weights to MLX safetensors.

Scores a moonshine-go sample against the Tier, Complexity, Setup Cost, Pillars, Industry, and Appeal rating schema and produces the Sample Rating markdown block for its README. Use when creating a new sample under samples/ or updating an existing sample's rating.

Transcribe local audio files or gs:// GCS objects to text using the moonshine-go CLI's "transcribe" command (Moonshine STT via libmoonshine). Use when the user asks to transcribe, caption, or get a text transcript of a .wav file (or other audio convertible to .wav), in this repo or any project that vendors this CLI.

Interact with Google Drive, Docs, and Sheets using the drivectl CLI. Use this skill when asked to list Drive files, download files, read/update Sheets, or create/export Google Docs.

Interact with Agent-to-Agent (A2A) protocol services from the command line. Use when you need to discover agent capabilities, send tasks, poll status, stream results, manage tasks, or authenticate with OAuth 2.1-protected A2A agents.

Utilizes the agentskills CLI tool to recursively scan local directories or public GitHub profiles for workspace rule conventions (GEMINI.md), parsing them with gemini-3.5-flash to aggregate capabilities and identify duplication.

Transcribes audio (.wav/.flac/.mp3/.m4a/.ogg) locally using Gemma 4's native audio encoder via the gemma-stt CLI. Use when asked to transcribe, caption, or extract text from speech/audio files, or about local/offline speech-to-text with Gemma models.