What does this skill do, and when should you use it?
This is a persona Agent Skill: once installed, the AI responds as Justin Sun in first person rather than paraphrasing his views. Its source base is unusually deep — statistical analysis of 14,418 English and 7,411 Chinese tweets, the full text of his autobiography, complete transcripts of his 155-episode Himalaya audio course, five long-form interviews (花总, LatePost, Bankless, UpOnly, the Blue Origin flight) and 42 investigative reports, all cross-validated. The output includes 1 meta-model plus 13 mental models, 19 decision heuristics, a fully data-driven expression DNA (emoji, high-frequency words, rhetorical patterns), six documented contradictions, and explicit honesty boundaries. It auto-switches between Chinese and English styles and includes a clean exit mechanism.
The skill is a single large SKILL.md prompt file — no scripts, no network calls. When triggered it: classifies the question (factual / framework / hybrid); runs a Sun-style 'assess scale → assess trend → assess timing' analysis of crypto projects, business moves or industry trends; answers in first person with parallel-structure aphorisms, grand analogies (Water Margin, poker, biological evolution) and a 🔥🚀🎉-flavored close; draws on 14 mental models (Increment vs Stock, Winner-Takes-All, cold-blooded→warm-blooded evolution, attention-as-mass E=mc², etc.) and 19 heuristics as analytical tools; states a one-time disclaimer on first activation, and returns to normal mode when the user says '退出' / 'exit'.
- A crypto operator who wants an aggressive 'winner-takes-all + incremental-thinking' stress test applied to a sector or project they are evaluating
- A founder who wants to simulate how a highly controversial, all-in-style serial entrepreneur would critique their fundraising and expansion plans
- A content creator studying Sun's persuasion patterns (parallel aphorisms, number-bombing, trend-riding, break-then-build) as marketing copywriting reference
- A researcher or journalist who wants a quick reference to his six documented words-vs-actions contradictions and a five-era evolution timeline
- A Chinese-speaking user interested in attention economics and memecoin logic who wants a punchy, aphorism-dense thinking partner
How do you install this skill?
- Persona output may be mistaken for Justin Sun's actual views; the disclaimer appears only once and misleading-output risk accumulates in long conversations — consider repeating it per session or before consequential decisions.
- Several mental models (e.g., Model 11, 13) rest on single sources and are independently unverified; do not use conclusions for investment or legal decisions.
- Inconsistencies between description and body (14 vs 1+13/14 models; 18 vs 19 heuristics) — treat the body as authoritative.
- Use of transcribed autobiography and paid audio course lacks formal authorization statements; copyright dispute risk exists.
- No versioning or maintenance commitment; corpus extends to 2026 with no freshness disclosure.
- Publisher is unverified; this is a static review — no code was executed and no output was reproduced.
- Local filesystem
With Claude Code, run in your terminal: npx skills add 0xquqi/sun-skill. To install manually, copy SKILL.md from the repo into your AI assistant's system prompt or custom instructions. The README states an MIT license (the repo description field says 'unknown' — defer to the repo).
git clone --depth 1 https://github.com/0xquqi/sun-skill.git ~/.claude/skills/sun-skillGenerated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Use Justin Sun's perspective
- What would Justin Sun think?
- Switch to Sun mode
- 用孙宇晨的视角
After installation, trigger it with phrases like “Use Justin Sun's perspective”, “What would Justin Sun think?”, “Switch to Sun mode”, or Chinese “用孙宇晨的视角” / “切换到孙哥”. Then ask your question normally — the AI will answer in Sun's first-person voice. To end the roleplay, say “退出”, “切回正常” or “exit character”.
What are this skill's strengths and limitations?
- Exceptional source coverage: 21,829 tweets with dual-account statistics, full autobiography, 155 audio episodes, five long interviews and a 269+ source index
- Rigorous method: every model and heuristic is tagged with sources and cross-validation counts; contradictions and evasive topics are documented honestly
- Data-driven expression DNA (emoji frequencies, word counts, rhetorical stats) grounds the persona, with bilingual style switching built in
- Pure prompt file: zero dependencies, zero cost — paste it into any client that supports system prompts
- It is a roleplay skill — output is Sun-style inference, not his actual views, and is a weak basis for real decisions
- Corpus ends around early 2026; later developments (litigation, markets, politics) are outside the framework
- No test suite or evaluation in the repo; persona fidelity depends on the host model, and no platform beyond Claude Code is claimed to be verified
- The material involves a highly controversial figure and contested claims (including dated events like the SEC settlement and Trump investment) — verify facts independently before relying on them
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Justin Sun Cognitive OS (Sun Skill) this page | 47 · Use with care | ★ 338 | 6mo ago | — |
| Recursive Research | 50 · Use with care | ★ 49 | 6mo ago | MIT |
| Claude Council | 47 · Use with care | ★ 97 | 3mo ago | — |
| Claude Council | 57 · Use with care | ★ 97 | 3mo ago | — |
| Minimalist Review | 54 · Use with care | ★ 11k | 6mo ago | — |
The README credits the Nuwa Skill (alchaincyf/nuwa-skill) as its structural inspiration and internally contrasts Sun's framework with a CZ-style persona (emoji fingerprints, language style, decision philosophy), suiting users who like the 'cognitive operating system of a person' format.
How did FollowSkills review this skill?
SKILL.md is a pure text role-play framework: no permissions, no code execution, no external calls, no sensitive data handling or destructive defaults; no red-line risk visible. Deducted for: impersonation of a living person carries misleading-output risk; the disclaimer is stated only once at first activation and not repeated; attribution of transcribed paid content (audio course, autobiography) rests on README goodwill rather than formal authorization; publisher unverified.
As a single-file prompt skill, instructions are self-consistent, triggers explicit, role rules coherent including an exit mechanism, and no runtime dependencies exist. Deducted for: static review cannot verify output quality; no tests or failure-feedback design for abnormal input; no behavior boundary for out-of-scope questions; claims data through 2026, creating temporal mismatch risk.
Trigger phrases are extensively listed in both languages, target scenario is clear (crypto/business thinking advisor), bilingual support is native with Chinese as core, and no overseas-service dependency affects mainland-China users. Deducted for: capability boundaries and non-fit disclosure are absent (e.g., not investment advice), and broad triggers may cause false invocations.
Well-layered information architecture (models, heuristics, DNA, contradictions, timeline), README covers install/usage, MIT license stated. Deducted for: description says 14 mental models + 19 heuristics while the body shows 1 meta + 13/14 models and inconsistent heuristic counts (18 vs 19); no versioning, changelog, maintenance responsibility or update path; license is repo-level, not confirmed in the skill itself.
The core task path is complete: classify-research-answer workflow, language fingerprints, and rebuttal templates support directly usable persona text, with clear marginal value given the extreme cost of assembling such corpus manually. Deducted for: static review cannot verify usable output; several models rest on single sources; accuracy unverified.
Rich citations (book chapters, podcast names, interview years, tweet statistics) and a claimed 269+ source index, plus explicit cross-validation of contradictions showing fact/inference separation awareness. Deducted for: quotations and statistics cannot be independently checked in static review, and no third-party execution evidence supports reproducibility.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →