A Claude Code skill for structuring catalyst-driven market research, Wisdom-of-the-Crowds signal assessment, and swing-trade decision discipline.
A decision-support framework for:
- evaluating the quality, independence, specificity, and trajectory of crowd signals
- structuring catalyst-driven swing trade theses with required invalidation conditions
- managing concentration risk with explicit thresholds
- making pre-earnings decisions with pre-written action plans
- separating long-horizon capital from short-term swing accounts
- planning exits toward hard liquidity deadlines
- publishing crowd signal analyses with mandatory financial disclaimers
The core function is to evaluate whether a crowd narrative is coherent and well-evidenced — not to tell users what to buy.
This is not a stock-picking bot, financial adviser, signal service, performance guarantee, buy score, risk score, expected-return model, or personalized investment recommendation.
The Crowd Signal Quality Score measures evidence convergence. It does not measure expected return, security safety, or whether a user should buy, sell, or hold anything.
See DISCLAIMER.md and docs/legal-risk-release-checklist.md.
- Retail investors who want structured decision discipline before deploying capital
- Traders who want pre-written exit and invalidation rules for every position
- Users who want to separate crowd signal quality from trade fit
- Anyone who wants to keep private portfolio context out of public skill logic
# 1. Copy the skill into your Claude skills directory
cp -r skills/catalyst-swing-intelligence/ ~/.claude/skills/
# 2. Copy and fill in your private personalization
cp skills/catalyst-swing-intelligence/PERSONALIZATION.example.md \
skills/catalyst-swing-intelligence/PERSONALIZATION.local.md
# 3. Edit PERSONALIZATION.local.md with your own accounts, risk limits, and rules
# 4. Never commit PERSONALIZATION.local.mdThis skill works best alongside two companion skills (not included here):
staged-position-entry— entry discipline, tranche sizing, blended cost basis, dry powder rulesthesis-break-protocol— exit rules when a thesis fails, three-strike rule, rationalization checklist
See docs/mode-decision-tree.md.
| User intent | Mode |
|---|---|
| "What is the crowd saying?" | crowd-scan |
| "Build a trade thesis." | thesis-build |
| "Should I hold through earnings?" | earnings-gate |
| "One position is too large." | concentration-check |
| "What is the macro regime?" | macro-first |
| "I have cash after a trim." | next-cycle |
| "There is a geopolitical event." | geopolitical-catalyst |
| "I'm second-guessing a trim." | regret-check |
| "Write a compliant social post." | publish-signal |
| "This is a long-term account." | long-horizon-account |
| "I have a hard liquidity deadline." | liquidation-plan |
See docs/crowd-signal-scoring.md.
The Crowd Signal Quality Score (0–100) measures evidence convergence:
- Signal volume, source independence, specificity, evidence quality, time acceleration, catalyst alignment, and dissent quality
- Penalties for meme/hype, crowding, price-movement, single-source, and loose baskets
- Signal trajectory classification: Emerging → Accelerating → Mainstreaming → Saturated → Fading
Crowd Signal Quality ≠ Security Risk ≠ Trade Decision
A high score means the signal is well-evidenced. It does not mean you should buy.
This repository contains only public reusable methodology.
Never commit:
- current holdings or cost basis
- personal or family names
- account values or balances
- liquidity deadlines
- tax details
- personalized financial plans
PERSONALIZATION.local.md
PERSONALIZATION.local.md is gitignored by default.
Use PERSONALIZATION.example.md as a template.
See CONTRIBUTING.md,
docs/contributor-evaluation-checklist.md,
and docs/legal-risk-release-checklist.md.
New users can run:
python tools/csi/csi.py wizardThis walks through the evidence-to-memory workflow without requiring users to memorize every command. Use --dry-run to preview the steps non-interactively.
This repo includes a minimal search-first reference implementation under tools/csi/.
It does not require paid APIs.
It can generate search queries, validate and import evidence, score evidence deterministically, write a markdown crowd signal report, and store observations in a local memory flywheel.
Use it when you want a low-cost local workflow or a deterministic scoring helper for the agent skill.
python tools/csi/csi.py queries "AI data center power scarcity"
python tools/csi/csi.py template --output evidence.csv
python tools/csi/csi.py score tools/csi/sample_evidence.csv
python tools/csi/csi.py report tools/csi/sample_evidence.csv --output report.md
python tools/csi/csi.py demoSee tools/csi/README.md and
docs/reference-implementation.md.
The reference implementation can store scored crowd-signal observations locally, attach later outcome reviews, generate monthly effectiveness reviews, and suggest updates to a local crowd-signal playbook.
This helps evaluate which crowd-signal patterns were useful over time.
It does not evaluate whether the skill picked winning securities.
The crowd-signal playbook is not an investment playbook and must not suggest purchases, investments, trades, position sizing, or buy/sell/hold actions.
python tools/csi/csi.py observe tools/csi/sample_evidence.csv --theme "AI infrastructure"
python tools/csi/csi.py list
python tools/csi/csi.py outcome SIGNAL_ID --usefulness useful --failure-mode none
python tools/csi/csi.py monthly-review --month 2026-05
python tools/csi/csi.py playbookThis project welcomes contributions that improve crowd-signal assessment, evidence quality, scoring consistency, documentation, and local operability.
The official repo does not accept changes that turn the project into a stock picker, signal service, financial advice tool, portfolio allocator, or recommendation engine.
Before opening a PR, read:
All PRs should preserve:
Crowd Signal Quality ≠ Security Risk ≠ Trade Decision
Crowd Signal Playbook ≠ Investment Playbook
Open source means anyone can fork the project. The official repository maintains stricter standards: tests must pass, non-advisory boundaries must remain intact, and private-data protections must not be weakened.
This project is usable today as a search-first, local, deterministic crowd-signal intelligence workflow. The current known gaps are:
- No built-in web search runner yet.
- No local dashboard yet.
- No large-scale real-world calibration yet.
See docs/known-gaps.md and docs/roadmap.md.
See docs/evaluation-tests.md for a manual test
suite to run after any change to SKILL.md.
See docs/testing-guide.md for the automated test
suite for the reference implementation.
This is a decision-support framework, not financial advice.
See DISCLAIMER.md for the full disclaimer.