01 / SCANNER
Find candidates
Daily pattern scans, opening-range breakout checks and filtered public symbol lists.
11 core workflows
A familiar day-trading workflow, rebuilt as independent skills and a small Python calculation layer.
02 / NEWS
Watch catalysts
Headline metadata, event tags, timestamps and optional SEC filing links.
03 / ENTRY
Confirm the bar
Closed-candle triggers with freshness checks; an unfinished candle cannot confirm.
04 / EXIT
Plan exits
ATR-based stop and target levels with assumptions shown explicitly.
05 / TIME MACHINE
Test scenarios
Historical path plus seeded OHLC bootstrap scenarios and limitations.
06 / RISK
Block unsafe proposals
Exact-order PASS/BLOCK checks, loss budget, exposure and order fingerprint.
07 / BUDGET
Size by risk
Whole-share estimates with fees, slippage and a notional cap.
08 / PAPER BROKER
Simulate only
A local ledger. No live broker integration, exchange routing or stored broker secrets.
09 / ALERTS
Set thresholds
One-shot local price alerts checked manually or from a user's own scheduler.
10 / ANALYST
Read reported data
Selected SEC XBRL metrics and recent filing metadata with period and unit context.
11 / DISCLOSURES
Track public records
House PTR index and source-PDF workflow; no inference of motives or trading edge.
Six research extensions
Market regime · walk-forward Strategy Lab · Portfolio Exposure · Data Quality · Trade Journal · Post-Trade Review.
Multi-model by design
Skills are plain Markdown, not vendor API calls. Native Agent Skills are packaged for ChatGPT, Codex, Claude Code, Grok CLI and Gemini CLI; a skills-only OpenAI plugin manifest is included for supported ChatGPT/Work plugin surfaces. Gemini Apps can import individual skills. Each host still applies its own execution and data-access limits.
Research flows forward. Risk stays separate.
Pattern detection is not a recommendation. Risk PASS is not an execution approval.
The overview is an original workflow animation, not a recording of live market performance.
Get running
Python toolkit
python -m venv .venv
# Activate the environment
pip install -e .
python -m unittest discover -s tests -v
marketpilot --help
marketpilot budget --equity 10000 \\
--risk-pct 0.5 --entry 100 --stop 98Install agent skills
python scripts/install_skills.py all
# all = Claude Code + shared .agents/skills
# dedicated choices: grok, gemini, codex, claude
# ChatGPT Chat/Work + Codex plugin surfaces: plugin.json
# Gemini Apps: upload one skill folder/ZIP
# Grok/Gemini CLI: use native skill foldersSee model compatibility for exact paths and limits.
Claims are earned, not written into a README
No performance promise. Included tests are offline and use deterministic synthetic fixtures. No return, win-rate, latency or accuracy benchmark is claimed.
Market data may be delayed, incomplete or rate-limited. Simulations are assumptions about historical samples, not predictions. MarketPilot AI is not financial advice and is not a substitute for regulated providers or real-time licensed feeds.