← ClaudeAtlas

forecast-workflowlisted

Use whenever someone needs a forecast, a probability or a launch estimate and is not studying a specific chapter. Seven gates, each producing an artifact the next command requires, so a person with no forecasting experience and an expert follow the same auditable path from question to scored result.
KarpelesPublishing/the-art-and-science-of-forecasting · ★ 0 · AI & Automation · score 72
Install: claude install-skill KarpelesPublishing/the-art-and-science-of-forecasting
# Forecast Workflow: the seven gates This is the front door of the library. It does not teach a method; it makes sure the right one is chosen and used honestly. Every gate produces a file, and the next command refuses to run without it, so no gate can be skipped, whatever the reader's experience. The chapter skills (`forecasting-chNN-*`) explain the methods; [all-chapters-forecasting](../all-chapters-forecasting/SKILL.md) maps them; [conventions.md](../all-chapters-forecasting/references/conventions.md) holds the shared rules. Commands run from the project root with `companion/.venv/bin/python companion/scripts/run.py`. The reader may not know any of these words. Explain each gate in one sentence before asking its questions. Never run a gate on the reader's behalf without telling them what it will do; never skip one because the reader is in a hurry. A forecast produced without the brief or the profile is not finished, and the report will say so. ## Gate 1: the brief (what number, for what decision) ```bash run.py brief --output work/brief.json --target "..." --units ... --decision "..." --horizon N --frequency monthly --as-of YYYY-MM-DD --outcome-date YYYY-MM-DD ``` The command prints the questions still unanswered. Ask them in that order, one at a time, in the reader's words, and rerun the command with the answers. It exits 0 only when the brief can support a forecast (target, units, decision, horizon, frequency, scoring date). The two questions readers most often canno