Yofine Skills
Five skills from one author: arXiv reports, technical diagrams, a report site, a design system library and agent guardrails
- 155 ★ on the repository
- Last commit
Extended access
Why this level
- `brief` and `design-hub` create projects with npm dependencies and start a local server
- `arxiv-paper-report` runs scripts that normalise an arXiv stream and build the report on your machine
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/yofine-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- `brief` and `design-hub` create projects with npm dependencies and start a local server
- `arxiv-paper-report` runs scripts that normalise an arXiv stream and build the report on your machine
This is third-party code. Review the repository files before installing.
What it does
The repository gathers five standalone skills. arxiv-paper-report turns a set of arXiv papers into an evidence-bounded view of a field: claims tied to sources, capability boundaries, comparable and non-comparable technical options, decisions and leading indicators. The output is strict JSON plus a bilingual standalone HTML report. blueprinter draws technical diagrams in a flat engineering blueprint style with HTML and CSS. brief sets up a React and Vite report site, builds standalone HTML analyses and publishes them into the site. design-hub runs a local design system library: it extracts a system from screenshots or a description, builds a safe mock and exports tokens, `DESIGN.md` and an agent-facing API document. By default the library lives in `~/design-hub`. ten-defeats stands apart: it guards an agent against ten failure modes in non-trivial work, such as action theater, tools standing in for judgment and declaring completion too early.
Who it is for. People who often ask an agent for analytic reports, architecture diagrams or design systems and want ready skills with examples.
Good fit when
- You want to know what recent arXiv papers on a narrow topic really show, with a report that links to sources
- An architecture diagram has to look like a neat engineering drawing
- Analyses of papers and code pile up in one place and your own report site publishes them
- Several interface screenshots have to become a design system handed to an agent as `DESIGN.md`
- An agent takes on a hard task and has to confirm the result before declaring it done
Not a fit when
- You want one narrow skill: without the `--skill` flag the other four install with it
- An arXiv report must rank approaches evaluated on different tasks: the skill refuses to compare them in such cases
- You need a report site on another stack: the brief template is built on React, Vite and Tailwind
Example request
Use arxiv-paper-report: study recent papers on proactive AI agents and tell me what is demonstrated, what remains unknown and what to validate nextLimitations
- The author does not describe installation: the generic `npx skills add yofine/skills` assumes a `SKILL.md` in each folder.
- The repository calls itself a personal collection, and the MIT license is stated only in the README text.
- The `brief` and `design-hub` skills create React and Vite projects with npm dependencies and start a local server.
- Each skill gets a short README entry: details sit in `SKILL.md` and, for `arxiv-paper-report`, in its own README with dependencies and validation commands.
- The last commit in the repository dates from August 2026.
How to disable. Delete the folders of the skills you no longer want from the agent's skills directory. Created projects stay on disk: delete the `~/design-hub` directory and the `brief` report site by hand, and stop a running local server in the terminal.
FAQ
How do I install yofine's skills in Claude Code?
The author gives no dedicated command. The generic route is `npx skills add yofine/skills`: each repository folder holds a `SKILL.md`, so the CLI finds five skills. For one skill add the `--skill` flag with its name.
What does an arxiv-paper-report contain?
A strict JSON research record and a standalone bilingual HTML report. It holds a direct answer, an epistemic ceiling, capability frontiers, a mechanism model labeled as source or analyst, comparable options, five-dimension maturity and leading indicators.
Where does design-hub keep design systems?
A new site is created in `~/design-hub` by default and ships with the MuleRun, Raft, QoderWork and Superset systems. New systems are appended as separate entries without rewriting existing ones.
What does blueprinter do?
It draws architecture diagrams, system diagrams and flowcharts as technical specification sheets in the Flat Engineering Blueprint style using HTML and CSS. The README example shows a Claude Code architecture sheet.
When does ten-defeats trigger?
In non-trivial work: planning, edits, debugging, verification, driving a browser or several tools. The skill asks for a stated problem before action, a cheap test of the riskiest assumption and confirmation of the outcome the user will see.
Discussion
No questions or reviews yet.
Related
Skillscafe3310's personal skill collection: spec-driven project work, reviews, reports, browser, 3D and video in one repository
Twenty-two standalone skills: Notion, Todoist, Tavily, Context7, image generation, converting documents to Markdown
Ten ShareAI Lab skills: source-level architecture research, review of past agent sessions, meeting coaching and article adaptation
A personal set of six skills: goal briefs for an agent, disk cleanup, AI news digests, project tidy-up, deep research and long-form writing
Twenty-seven Chinese-language skills for development, thinking, learning, writing, visuals, branding and working with agents
Seven skills for research, HTML explainers and source citation, installable all at once or one at a time through npx.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Yofine Skills (this card) | no | 155 | |
| arXiv Review Paper Harness | no | 478 | guides an agent to a review paper for arXiv in LaTeX with every reference checked; it builds no field-state report |
| DESIGN.md | no | 9 | a plugin and CLI that create DESIGN.md from the spec's templates, without a systems library or mocks |
README in short
The README presents the repository as a personal collection, shows the folder layout and a table of five skills. arxiv-paper-report, brief and design-hub get descriptions with screenshots and sample requests, and the MIT license is stated at the end.
README badge
Original SKILL.md
---
name: ten-defeats
description: "Use when Codex is asked to plan, design, implement, modify, refactor, debug, test, review, verify, or complete non-trivial software work; use Computer Use or operate a browser, desktop UI, or live app; coordinate multiple tools; or delegate to agents. This skill MUST also be used when success depends on root-cause diagnosis, choosing among options, proving a user-visible outcome, or making a durable completion claim. Do not use for simple facts, translation, or trivial text-only edits."
---
# Ten Defeats
Core rule:
> Let judgment govern tools. Never use tools to manufacture the appearance of
> judgment.
Use only the relevant guards. Never perform or print a ten-item ritual. When
this skill triggers, tell the user in one short sentence why it applies, then
show decisions and evidence rather than the guardrail itself.
## The ten defeats
1. **Intent Defeat** - State the real problem before acting. Do not compensate
for weak framing with a larger plan.
2. **Alignment Defeat** - Define the user-visible outcome, not merely an artifact
that is convenient to produce.
3. **Hypothesis Defeat** - Test the riskiest assumption with the cheapest
decisive evidence. Test count and coverage are not substitutes for thought.
4. **Tooling Defeat** - Form the design before using Computer Use. Every tool
call must answer a question or advance the agreed outcome.
5. **Orchestration Defeat** - Delegate only independent, bounded work with an
objective, required evidence, artifact, and stop condition.
6. **Root-Cause Defeat** - Establish a causal model before fixing. Stop and
re-diagnose before adding a second symptom workaround.
7. **Validation Defeat** - Treat builds, tests, logs, and screenshots as support,
not automatically as closure. Verify the real primary path.
8. **Clarity Defeat** - Separate fact, inference, and unknown. Say no more than
the evidence supports.
9. **Decision Defeat** - Recommend a default with reasons. Ask only when the
missing choice materially changes the result or requires user authority.
10. **Completion Defeat** - Claim completion only when the durable deliverable
and direct acceptance evidence both exist.
## Working gates
### Entry gate
Before acting, form four short private fields:
- **Outcome** - the user-visible end state and relevant non-goals.
- **Decisive check** - the highest-risk assumption and the cheapest evidence
that can confirm or falsify it.
- **Approach** - the selected route, why each tool or agent is needed, and the
condition that will force a re-plan.