| name | agentmatrix-alpha-pipeline |
| description | Run the AgentMatrix alpha pipeline from alpha_resource.md through internal validation, strategy packaging, external simulation, and live-readiness gates. |
AgentMatrix Alpha Pipeline Skill
Use this skill when the task is to run or review the full alpha research pipeline in agentmatrix-research.
Stages
- Feeder produces
alpha_resource.md.
- Factor research normalizes
FactorResearchSpec, implements the factor, and runs internal validation.
- Validated factors are promoted to
strategy_candidate only when data quality and red-flag checks are acceptable.
- Strategy engine builds target weights with the same factor signal path used in research.
- Backtest adapter packages GM, PTrade, or QMT external simulation files.
- Returned terminal simulation evidence is parsed and attached before any
live_ready promotion.
Commands
Check amazingdata:
python -m research_core.factor_lab.cli check-amazingdata
Run internal validation:
python -m research_core.factor_lab.cli run-factor-research --factor-set wq101 --data-source amazingdata --start 2023-01-01 --end 2025-12-31 --universe csi800 --max-symbols 300
Build strategy signals:
python -m research_core.strategy_engine.cli build-alpha-strategy --validated-run runtime/factor_lab/jobs/<job_id>.json --rebalance-frequency daily --top-n 50
Package external simulation:
python -m research_core.backtest_adapter.cli package-external-sim --engine gm --strategy <strategy_id> --signal-path runtime/strategy_engine/<strategy_id>/target_weights.csv --start 2023-01-01 --end 2025-12-31
Proof Rules
- Internal validation is necessary but not sufficient for live trading.
- Never promote to
live_ready without external simulation evidence and explicit human approval.
- Never auto-flip factor direction to improve metrics.
- For amazingdata, named universes must be resolved and intersected with point-in-time ClickHouse listing status from
ods_security_status_daily; use --universe all only when index membership is intentionally not enforced.
- Warmup rows are allowed for factor computation, but validation evidence must be trimmed to the requested
[start, end] window.
- Multi-year external simulations require multi-date strategy signals; use
--as-of only for a single-snapshot smoke export.
- Stop on failed data-quality status, missing required fields, empty factor coverage, or unresolved lookahead risk.
