{
  "markdown": "# Skills\n\n[![NVIDIA SkillSpector audit](https://img.shields.io/github/actions/workflow/status/wquguru/skills/security-audit.yml?branch=main&logo=nvidia&label=NVIDIA%20SkillSpector%20audit)](https://github.com/wquguru/skills/actions/workflows/security-audit.yml)\n[![AI review by Claude](https://img.shields.io/badge/AI%20review-Claude-d97757?logo=claude)](https://github.com/anthropics/claude-code-security-review)\n\nA small collection of practical Agent Skills.\n\n> **Public monorepo.** Anyone can clone this repo, run `just add`, or install\n> individual skills with `npx skills add`. It does not depend on any private\n> tooling. The install loop links into Claude Code, Codex, and Grok skill dirs.\n\n**Catalog:** self-authored skills live under [`skills/`](./skills/) (see also the\n[GitHub Pages registry](https://wquguru.github.io/skills/)). Highlighted ones\nbelow; the full set is every folder with a `SKILL.md`. Third-party modules are\ndeclared in [`external.yml`](./external.yml) and materialised by `just vendor`.\n\n## Skills\n\n### `claude-5-best-practice`\n\n> Renamed from `fable-5-best-practice` in July 2026, when Claude Opus 5 shipped\n> near-Fable capability at half the price and the routing advice stopped being\n> about one model. Reinstall under the new name.\n\nOptimizes the cost of an *accepted result* across the Claude 5 generation —\nHaiku 4.5 / Sonnet 5 / Opus 5 / Fable 5 — rather than the price of a token. It\ntreats Opus 5 as the default lead, makes Fable 5 something you justify instead\nof assume, and prefers climbing the effort ladder (`low` → `max`) on Opus 5\nbefore escalating a tier. It also walks a diagnose-before-escalating loop for\nfailures, routes subagents with pinned worker models, covers Claude 5 API\nbehavior (adaptive thinking, effort and prompt caching, refusal stop reasons,\nfallback, ZDR and Priority Tier constraints), and keeps progress claims tied to\nreal tool evidence. Volatile facts such as pricing, retention, and beta headers\nlive in a dated `references/evidence-notes.md` provenance snapshot.\n\nThe skill is especially useful when migrating older prompts or skills forward:\nit calls out over-constrained legacy prompting, the Opus 4.8 → Opus 5 thinking\nand effort breaks, refusal risks, verifier agents, long-running run design, and\nlong-run communication style.\n\nInstall:\n\n```bash\nnpx skills add https://github.com/wquguru/skills --skill claude-5-best-practice\n```\n\nAsk for help with prompts like:\n\n- \"Does this task actually need Fable 5, or is Opus 5 enough?\"\n- \"Help me write an Opus 5 prompt for a multi-day coding task.\"\n- \"Review this old skill before migrating to Claude 5.\"\n- \"My run missed the acceptance criteria — raise effort or change model?\"\n\n### `gpt-5-6-best-practice`\n\nA routing and evaluation overlay for GPT-5.6 in Codex. It optimizes the cost of\nan accepted result — not the price of a token — across the Sol, Terra, and Luna\ncapability tiers: choosing the initial tier and reasoning-effort lane,\ndiagnosing failures before escalating, keeping prompts and worker packets lean,\nusing subagents economically, and evaluating routes on real work. Benchmark\nprovenance and confidence levels live in `references/evidence-notes.md`, and\n`references/prompt-patterns.md` provides reusable task-prompt and worker-packet\ntemplates.\n\nInstall:\n\n```bash\nnpx skills add https://github.com/wquguru/skills --skill gpt-5-6-best-practice\n```\n\nAsk for help with prompts like:\n\n- \"Which GPT-5.6 tier and effort should this task start on?\"\n- \"This Codex run failed — raise effort, or move from Terra to Sol?\"\n- \"Is Ultra worth it for this task, or should one agent handle it?\"\n- \"Design a worker packet for this subagent.\"\n\n### `english-swe-daily`\n\nA practical English coach for everyday engineering communication. Built for\ndevelopers who can already work in English but want to sound less textbook-like\nand more like a real teammate in standups, Slack threads, 1:1s, code reviews,\nand meetings.\n\nMany engineers do not struggle with grammar — they struggle with sounding\nnatural. This skill teaches the kind of English people actually use in software\nteams, with a strong focus on natural phrasing, collaboration tone, and\nday-to-day workplace situations.\n\nInstall:\n\n```bash\nnpx skills add https://github.com/wquguru/skills --skill english-swe-daily\n```\n\nAsk for help with prompts like:\n\n- \"Teach me English for standups.\"\n- \"Help me sound more natural in Slack.\"\n- \"Practice English for code reviews.\"\n- \"How do I say this more naturally at work?\"\n- \"Give me today's SWE English session.\"\n\n### `pi-setup`\n\nConfigures Pi Agent (`@earendil-works/pi-coding-agent`) with DeepSeek V4\n(built-in provider) and Ant-Ling Ring-2.6-1T (custom OpenAI-compatible\nprovider), plus a curated extension set, while automatically working around 7\nknown pitfalls.\n\nIt detects before it acts: it probes whether pi is installed, whether config\nalready exists, and where keys live before changing anything. Keys are never\nwritten in plaintext — pi config only stores lazy `!shell` commands and the key\nstays in the user's original file. Model selection, default model, default\nthinking tier, and the extension set are all confirmed interactively.\n\nInstall:\n\n```bash\nnpx skills add https://github.com/wquguru/skills --skill pi-setup\n```\n\nTrigger it in Claude Code by saying \"配置 pi\", \"setup pi agent\", \"把\nring/deepseek 配到 pi\", or run `/pi-setup` manually.\n\nPrerequisites:\n\n- Node.js / npm (to install pi)\n- A DeepSeek key (in `~/.deepseek` or an environment variable)\n- An Ant-Ling/Ling key (`LING_API_KEY`, in a shell rc or environment\n  variable) — only needed when enabling Ring\n\n### `free-my-cpu`\n\nSafely audits CPU, load average, I/O wait, and Docker container pressure on\nLinux hosts, especially production-like servers over SSH. It follows the same\nread-only-first posture as `free-my-disk`: collect repeated samples, separate\nCPU-bound work from I/O pressure, inspect hot containers and logs, then classify\npossible fixes by impact before changing anything.\n\nInstall:\n\n```bash\nnpx skills add https://github.com/wquguru/skills --skill free-my-cpu\n```\n\nAsk for help with prompts like:\n\n- \"Use free-my-cpu to inspect this remote Docker host.\"\n- \"Why is this Docker host load average so high?\"\n- \"Find which container is burning CPU.\"\n- \"Use this Grafana URL and the last 6 hours to check whether monitoring queries are causing load.\"\n\n### `youtube-toolkit`\n\nDownloads YouTube videos with `yt-dlp` and post-processes them with `ffmpeg`:\nfetch best-quality streams, convert containers (e.g. AV1/webm → H.264 mp4),\nextract audio, and burn in translated subtitles with a translator watermark.\n\nIt is macOS-first and bakes in the gotchas learned the hard way — chiefly that\nHomebrew's lean `ffmpeg` ships without libass/drawtext (so subtitle burn-in\nneeds `ffmpeg-full`), that YouTube auto-captions are overlapping rolling cues\nthat must be re-segmented before they read cleanly, and that styling via an ASS\nfile avoids the `force_style` comma-parsing trap. Two helper scripts do the\ndeterministic work: `install-tools.sh` and `burn-subs.sh`.\n\nInstall:\n\n```bash\nnpx skills add https://github.com/wquguru/skills --skill youtube-toolkit\n```\n\nAsk for help with prompts like:\n\n- \"Download this YouTube video to ~/Downloads.\"\n- \"Convert it to mp4.\"\n- \"Add Chinese subtitles and a @handle 翻译 watermark.\"\n\nPrerequisites:\n\n- Homebrew (macOS) or apt/pip (Linux)\n- `yt-dlp` and a libass-enabled `ffmpeg` — `install-tools.sh` sets both up\n\n## Local Development\n\nFor local hacking, the `Justfile` is **self-contained**: it symlinks every skill\nunder `skills/` **and** `external/` into Claude Code (`~/.claude/skills/`),\nCodex (`~/.codex/skills/`), and Grok (`~/.grok/skills/`), so edits take effect\nimmediately without re-publishing. No private sibling repo is required.\n\n```bash\njust add      # symlink all skills (skills/ + external/) into agent dirs\njust remove   # remove only the symlinks that point back into this repo\njust status   # show install status for each skill in all destinations\njust vendor   # pull third-party skills declared in external.yml into external/\njust          # list available recipes\n```\n\n`add` is idempotent and never clobbers: an existing real directory or a symlink\nfrom another install tree is left untouched and reported as skipped\n(`foreign symlink`). `remove` only removes symlinks that point back into this\nrepo. After adding a new skill folder, re-run `just add`.\n\n## External (third-party) skills\n\n`skills/` holds the skills authored here (and published via `npx skills add`).\nThird-party skills from other repos are **not** copied into git — they are\nvendored separately so this public repo stays purely self-written and free of\nforeign licenses:\n\n1. Declare sources in `external.yml` — a repo, a `ref` (branch/tag/commit), and\n   the sub-paths to pull. Each dir's basename, prefixed with `3rd-`, becomes the\n   skill name (e.g. `dashboarding` → `3rd-dashboarding`).\n2. `just vendor` shallow/sparse-clones each source, copies the named skill dirs\n   into `external/` (git-ignored) under their prefixed names, rewrites each\n   `SKILL.md` frontmatter `name` to match, and records the exact upstream commit\n   SHA in `external.lock`.\n3. `just add` symlinks them alongside the self-written skills.\n\nOnly `external.yml` + `external.lock` are committed; the vendored code is not.\nRe-run `just vendor` any time to refresh (or after pinning a `ref`). A name that\nwould collide with a self-written `skills/<name>` is skipped; entries removed\nfrom `external.yml` are pruned from `external/` on the next `just vendor`.\n\n## Repository Structure\n\n```text\nJustfile          # add / remove / status / vendor (self-contained)\nexternal.yml      # third-party skill sources (repo + ref + sub-paths)\nexternal.lock     # resolved upstream commit SHAs (auto-generated)\nscripts/\n  vendor.sh       # pulls external.yml sources into external/\nskills/           # self-written skills (published)\n  claude-5-best-practice/\n  gpt-5-6-best-practice/\n  english-swe-daily/\n  …\nexternal/         # vendored third-party skills (git-ignored; just vendor)\nindex.html        # public site\n```\n\n## License\n\nSelf-authored skills under `skills/` are intended for public use via this repo\nand `npx skills add`. Third-party code under `external/` is **not** redistributed\nin git — only `external.yml` / `external.lock` are tracked; run `just vendor` to\nfetch upstream under their own licenses.\n",
  "bytes": 10402,
  "sha": "815f5a4ba762ae582bc6ab956ed525ceb071d8844eeb4c680eb1c13d1ac71b86",
  "repo_slug": "wquguru/skills",
  "fonte": "repo",
  "truncated": false,
  "api": "https://api.agentalog.com/api/listings/skl_wquguru_skills_fable5_best_practice_7d3b36ae/readme"
}