Serenity Skill
Ranks supply-chain layers before tickers, with graded evidence and kill switches
Test report
- Verdict
- Tested · Works
- Score
- Tested
- Jul 16, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · a139f33
Ran Test 1 from the built-in evals/test-cases.md: 'Deeply research A-share AI semiconductor industry chain, find 5 priority stocks' — first without skill, then strictly according to the body. Baseline response produced a familiar list of hyped tickers (寒武纪, 中芯国际, 海光) without layers, without strength of evidence, and without conditions for refutation — 0 out of 10 skill's own criteria; the skill's version yielded 10 out of 10: first, layer ranking (HBM/memory interconnect → advanced packaging/CMP → consumables), explicit demotion of the most popular layer (pure computing chips) with justification, strong/medium/weak labels on each thesis, a section 'what will prove the conclusion incorrect', and A-share-specific verification paths (互动易, 问询函, 招投标, 环评). The serenity_scorecard.py script was actually executed: weights sum exactly to 100, on my input it yielded 81.0 − 12.0 = 69.0 'Worth tracking', which matched my independent recalculation, and a rating of 99 was correctly rejected with ValueError. Stage 0 clean: read the full body + both Python scripts — only local JSON parsing, no network calls, no secret reading, no hidden instructions; SHA256.txt verifies for 24 out of 26 files (only README.md and README.zh-CN.md differ — manifest is outdated on docs, all executable and instructive content matches). I did not connect live market data, so the quality of the stock ranking itself was not checked — only the methodology was checked, and the skill itself correctly marked the run as initial pass.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 5/5
- Triggers reliably 5/5
- Output vs. baseline 9/10
- Docs & honesty 4/5
What Serenity Skill does
Turns an investment agent into a supply-chain bottleneck hunter: maps a theme into value-chain layers, finds the scarce layer, then names companies with graded evidence and failure conditions. Triggers on deep research requests for A-share/HK/US themes, value-chain mapping, or thesis challenges.
How to install Serenity Skill
git clone https://github.com/muxuuu/serenity-skill.git
mkdir -p ~/.claude/skills && cp -r serenity-skill ~/.claude/skills/serenity-skill
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Serenity Skill
-
/serenity-skillRanks supply-chain layers before tickers, with graded evidence and kill switches
It also activates on plain-language prompts like these:
-
Find the scarcest layer in the AI semiconductor supply chain, not just tickers -
Research A-share stocks for this theme with graded evidence, not hype -
Challenge my investment thesis and give me kill conditions for each pick
Frequently asked questions
- Is the Serenity Skill skill free?
- Yes. The skill itself is free from muxuuu/serenity-skill. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Serenity Skill work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 16, 2026. Verdict: Tested · Works. Ran Test 1 from the built-in evals/test-cases.md: 'Deeply research A-share AI semiconductor industry chain, find 5 priority stocks' — first without skill, then strictly according to the body. Baseline response produced a familiar list of hyped tickers (寒武纪, 中芯国际, 海光) without layers, without strength of evidence, and without conditions for refutation — 0 out of 10 skill's own criteria; the skill's version yielded 10 out of 10: first, layer ranking (HBM/memory interconnect → advanced packaging/CMP → consumables), explicit demotion of the most popular layer (pure computing chips) with justification, strong/medium/weak labels on each thesis, a section 'what will prove the conclusion incorrect', and A-share-specific verification paths (互动易, 问询函, 招投标, 环评). The serenity_scorecard.py script was actually executed: weights sum exactly to 100, on my input it yielded 81.0 − 12.0 = 69.0 'Worth tracking', which matched my independent recalculation, and a rating of 99 was correctly rejected with ValueError. Stage 0 clean: read the full body + both Python scripts — only local JSON parsing, no network calls, no secret reading, no hidden instructions; SHA256.txt verifies for 24 out of 26 files (only README.md and README.zh-CN.md differ — manifest is outdated on docs, all executable and instructive content matches). I did not connect live market data, so the quality of the stock ranking itself was not checked — only the methodology was checked, and the skill itself correctly marked the run as initial pass.
- What is the Serenity Skill SkillProof Score?
- 9.2/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 9/10, docs & honesty 4/5.
- How do I install Serenity Skill?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Serenity Skill with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.