North Star Investigation

Use this skill for the investigation workflow where quantitative reasoning and fastest-learning bets are the main job.

When to use

• The task is an optimization or root-cause hunt where the best move is not yet obvious.
• The user wants a Commander's Intent style investigation doc with ranked hypotheses and brutal tests.
• Math, budgets, effect sizes, or measured-vs-assumed distinctions materially affect the investigation quality.

When not to use

• The task is a straightforward bug fix with a normal bug doc. Use bugs-flow.
• The task is planned feature or architecture delivery. Use arch-plan, arch-mini-plan, or lilarch.
• The user wants a generic open-ended goal loop without the quant-heavy investigation frame. Use goal-loop.

Non-negotiables