quantmind-br

Sincroniza o repositório omo-profiler com o upstream oh-my-openagent usando o clone local canônico em ~/dev/oh-my-openagent (branch dev). Compara o schema JSON embarcado (internal/schema/schema.json) contra o artefato upstream, propaga mudanças de schema para internal/config/types.go, os wizard steps, testes e AGENTS.md, detecta drift de código desde o último anchor (internal/schema/.upstream-sha) mesmo quando o schema não mudou, valida com build/test/lint e paridade de hash entre os 3 arquivos de schema, e só então persiste o novo anchor. Use sempre que o usuário pedir para atualizar/sincronizar o omo-profiler com o upstream, verificar se o schema ou os tipos Go estão desatualizados/com drift, checar novidades do oh-my-openagent, revisar campos de config novos/removidos, ou invocar /update-omo-profiler. Keywords: sync schema, upstream sync, schema drift, oh-my-openagent, oh-my-opencode.schema.json, .upstream-sha, wizard fields.

Refactor and modernize an EXISTING TUI's UI/UX by reading its codebase, reverse-engineering the current design, diagnosing design smells, and producing an optimized target design plus a file-level refactor plan. Unlike tui-design (starts from a spec) this starts from real code; unlike tui-validator (only audits) this redesigns. Does NOT modify source code — outputs an actionable migration plan. Use whenever the user asks to refactor, redesign, modernize, overhaul, improve, clean up, standardize, or optimize a terminal UI/CLI app that already has code. Triggers include "refactor this TUI", "redesign my CLI UI", "modernize this terminal app", "melhorar UX da TUI", "refatorar TUI", "redesenhar CLI", "modernizar interface terminal", and "/tui-refactor". Prefer over tui-design when codebase exists; prefer over tui-validator when the user wants design improvement rather than only bugs/findings.

Audit a terminal user interface (TUI) by driving it inside a tmux session, capturing both text and pixel screenshots (live Wayland or headless Chromium fallback), probing every keybinding, hammering it with special characters and resize events, and producing a markdown diagnostic that lists broken bindings, shortcut conflicts, dropped inputs, rendering glitches and visual degradation. Use this skill whenever the user asks to validate, audit, smoke-test, QA, fuzz, regression-check or "find bugs in" a terminal/CLI UI — even when they don't name tmux, grim, or screenshots — including phrases like "test this TUI", "check keybindings", "see if the layout breaks", "review the design of my CLI tool", "does this TUI render correctly", "find shortcut conflicts" or "validate the keyboard shortcuts of <something>". Strongly prefer this skill over ad-hoc tmux scripting whenever the target is interactive (Bubble Tea, Textual, blessed, ratatui, ncurses, etc.).

Internal helper contract for calling the codex-companion runtime from pi-codex

Internal guidance for presenting Codex helper output back to the user

Internal guidance for composing Codex and GPT-5.4 prompts for coding, review, diagnosis, and research tasks inside the pi-codex package

Optimize markdown docs in a folder for AI agent consumption — Obsidian-flavored markdown, token-dense rewrite, parallel sub-agent dispatch via the pi-native `Agent` tool (tintinweb/pi-subagents extension), 100% content-integrity preserved. Use this skill whenever the user asks to format, optimize, compress, densify, AI-optimize, or rewrite a documentation folder previously produced by `repodocs-go` or the `organize-docs-skill` (folder containing `metadata.json` + numbered `nnn-*.md` files + `000-index.md`). Trigger on phrases like "format docs", "optimize docs for AI", "compress these docs", "obsidian format", "token dense", "format-docs", "ai-friendly docs", "library-docs format", "rewrite docs for agents", or any equivalent rewrite-for-density request. Runs AFTER `organize-docs-skill`. Requires the `pi-subagents` extension and the `format-docs-optimizer` + `format-docs-verifier` custom agents in `.pi/agents/`.

Organize a documentation folder produced by `repodocs-go` (or any crawler that emits `metadata.json` + `.md` files) into deterministic, lowercase, sequentially-numbered files plus a generated `000-index.md` optimized for AI agent consumption. Make sure to use this skill whenever the user asks to organize, normalize, renumber, sort, index, or clean up a documentation folder, or works with a folder containing `metadata.json` and bulk `.md` documents — even when phrased as "organize these docs", "sort the markdown files", "build me a docs index", "renumber this docs dump", "normalize the filenames", or any equivalent. Trigger on phrases like "library-docs", "metadata.json", "docs folder", "crawler output", "documentation index", "renumber files".

llama.cpp - C/C++ framework for LLM inference with GGUF models. Use when running, building, or optimizing LLM inference locally with llama.cpp. Keywords: llama.cpp, gguf, llm, inference, quantization, cuda, metal, vulkan, rocm, hip, opencl, multimodal, function-calling, speculative-decoding, build, cmake.