jokull/jokull-skills

最近记录的来源活动: 2026年10月7日 10:56

SkillsMP 收录数据更新: 2026年11月2日 18:02

已收集 skills: 9

GitHub 星标: 0
GitHub Forks: 0

用 AI 助手安装

复制这段 Prompt，发给你正在使用的 AI 助手。

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/jokull/jokull-skills.

仓库 README

仓库 README 来源：README.md

查看 README 源文件

展开完整 README

# jokull-skills Agent skills I actually use, in [Agent Skills](https://agentskills.io) format — portable across Claude Code, Codex, and other harnesses that read `~/.agents/skills/`. | Skill | What it's for | |---|---| | [`agent-vocabulary-coach`](skills/agent-vocabulary-coach/SKILL.md) | Incrementally expand a developer's vocabulary into precise, pragmatic engineering language that improves how they plan work and steer coding agents. | | [`roundtable`](skills/roundtable/SKILL.md) | Adversarial council review of a plan or spec. Independent specialist seats, run in parallel, synthesized into a patched artifact. | | [`sandbox-builder`](skills/sandbox-builder/SKILL.md) | Design and build an agent sandbox for a monorepo: a disposable per-worktree runtime where agents get a browser, seeded scenarios, x-ray visibility, and honest evidence. | | [`blender-product-render`](skills/blender-product-render/SKILL.md) | Product renders and 3D asset exports by scripting Blender headlessly — the `.blend` is an output, not a source. | | [`mole`](skills/mole/SKILL.md) | Judgment and memory on top of the [`mo`](https://github.com/tw93/mole) Mac cleanup CLI. Dry-run first, curate, remember the verdicts. | | [`message-local-codex`](skills/message-local-codex/SKILL.md) | Find and message local Codex sessions through Superset workspaces or session IDs, with return addresses and bounded replies. | | [`migrate-to-cf`](skills/migrate-to-cf/SKILL.md) | Move a Workers project from Wrangler config to the `cf` CLI and `cloudflare.config.ts`: codemod, parity table, dry-run verification, nothing deployed. | | [`memory-retro`](skills/memory-retro/SKILL.md) | A retro for what the agent remembers: check saved memories against the environment, fix what evidence settles, and reconcile conflicts with you one decision at a time. | | [`blitz`](skills/blitz/SKILL.md) | Clear a backlog in one push: inventory open PRs and issues, grill for product decisions, decide the engineering ones, then land worker output as a few large roll-up PRs. | ## Install ```bash git clone https://github.com/jokull/jokull-skills.git cd jokull-skills ./scripts/link-skills.sh ``` Skills are symlinked into `~/.agents/skills/` and `~/.claude/skills/`, so this repo stays the source of truth — a `git pull` updates the installed skills, and edits are live. ## Notes **`roundtable`** pairs with `/grill-me`: grill your idea into existence, then run it through the roundtable to bulletproof it. Descended from the `roundtable-review` skill I wrote in a work monorepo; the change here is that seats run as genuinely independent parallel subagents instead of sequential passes in one context, so the critiques can't anchor on each other. **`sandbox-builder`** is the [agent sandbox blueprint](https://md.solberg.is/2fbf166879b40940cd598565dc0a498a) restructured as a skill: the workflow and judgment in `SKILL.md`, the architecture, integration policy, and operating lessons in `references/`, loaded only when the phase needs them. The lessons come from running one of these in a real monorepo for a while. **`mole`** keeps its memory at `~/.local/state/mole-skill/memory.md`, outside the skill directory on purpose — the skill is shareable, but the verdicts name one machine's repos and databases. It seeds from a template on first run. ## Conventions See [CLAUDE.md](CLAUDE.md).

正在加载仓库 README…

这个仓库中的 skills

软件开发工程师平面设计师网络与计算机系统管理员项目管理专家

4 个职业分类 · 已分类 56%