agentbull/ankole/research-for-planning

Research for Planning

This Skill is for researching the environment around a controllable operating, product, market-entry, or launch decision. It helps a planner compare feasible action combinations. It is not a generic action-advice or plan-writing Skill, and it does not replace ordinary search.

Activate it only when both conditions hold:

1. The request needs substantial, dedicated external information gathering,

comparison, or synthesis. A quick lookup, one current value, or a supplied collection of documents is not enough.

1. The findings will choose or change a controllable action, strategy,

resource commitment, sequencing, or material constraint.

Start from the decision and the actions that could change. Identify only the unknowns whose answers would change that choice. Use focused sources and stop when the evidence is sufficient to select a route or define a bounded test. Do not keep researching to prove every premise, fill a report, or reach a quantified probability.

The useful output is an evidence-backed handoff to the planner: recommendations or comparisons, decision-relevant constraints, assumptions, source links, and the observation that would change the choice. If the user also asks for a complete plan, continue into that plan after the research; the plan request is not why this Skill was activated.

Keep these boundaries:

• Use create-deep-research when the research question is what is true in the

external world, how it may develop, whether a claim holds, or what outcome is likely. This includes stock buy/hold questions, price or demand forecasts, competitive outcomes, and risk estimates: the requested action does not change the research object.

• This Skill can estimate costs and map current constraints, but it does not

predict target-market growth, a stock price, demand, or another future outcome. Split those questions to create-deep-research.

• Do not activate either research Skill merely to write a schedule from given

material, discuss strategy, polish or review an existing plan, or collect more sources in a requested PDF.

• A mixed request can split its subquestions by purpose: research the

decision-changing action questions here, use deep research for external-world questions, then return to the requested action.

• When the investigation is small, search directly. Use ordinary background

work when it helps with duration, not as a new routing category.

Do not take over brainstorming for an explicit exploration request or proposal-review for review of an existing proposal.