strategy-generationlisted

# Strategy Generation A strategy that exists only as prose is a mood. This skill converts ideas into specs that machines can refuse, code that gates can check, and claims that survive the validation pipeline — or die there honestly. ## The pipeline (six stages, three gates) ``` IDEA → SPEC ──gate 1──→ CODE → RETURNS ──gate 2──→ RISK ──gate 3──→ PAPER │ │ │ │ validate_strategy_spec │ backtest- │ risk-assessment │ (this skill's script) │ validation │ profile ``` 1. **DATA** — establish market, universe, frequency, history budget, and whether the data pipeline is PIT-reviewed (`point-in-time-research`). 2. **SPEC** — fill the contract in `references/spec-schema.md`: named archetype, testable entry/exit logic, explicit parameters, at least one hard risk control, a cost assumption in bps, and `variants_tried`. 3. **Gate 1** — `python scripts/validate_strategy_spec.py SPEC.json`. FAIL means fix the spec; do not write code around a broken spec. 4. **CODE** — implement exactly the spec. Parameters come from `signal.parameters`, not from numbers invented mid-implementation. If implementation forces a change, update the spec and re-run gate 1. 5. **Gate 2** — produce the returns series and run it through `backtest-validation` with `--trials` = the spec's `variants_tried` (updated if more variants were tried during coding — honesty compounds). A FAIL verdict is a result: re