--- name: research-before-asking description: Research and carry decisions before asking the user. Use for ambiguous requirements, missing specifications, ‘which approach’ or ‘should I use X or Y’ choices, or taste, priority, private facts, credentials, or authorization that remain after proportionate research. Present viable options, tradeoffs, a strong recommendation, and its flip condition. Not for verifying claims, responding to an ungrounded unknown, or Post.Build.Ship approval and gate policy; use thinking-is-not-knowing or i-dont-know-is-the-answer. --- # Research Before Asking ## Start From First Principles Before asking the user, establish: 1. the objective the answer must serve; 2. the constraints and authorization already in force; 3. the failure modes each plausible choice must prevent; and 4. the facts discoverable from the repository, relevant history, documentation, commands, or other authoritative sources. Separate discoverable facts from judgment. Research does not mean collecting everything available; it means finding enough relevant evidence to choose, carry, or precisely frame the remaining decision. ## Use One Of Three Outcomes ### Proceed When evidence yields one outcome and that outcome is already authorized, state the finding and proceed. Do not ask the user to repeat research, choose a foregone conclusion, or approve an action they have already authorized. ### Carry A Material Choice When more than one viable outcome remains, do the decision work before asking. Present: - the viable options; - their material tradeoffs; - a strong recommendation tied to the objective and evidence; and - the flip condition: the fact, priority, or constraint that would make another option better. Ask only for the judgment or authority that remains. A menu without analysis transfers the work to the user; a recommendation without a flip condition hides what could change it. ### Ask The Smallest Question Ask directly when the remaining input is genuinely non-discoverable: taste, priority, a private fact, credentials, or authority. State what was checked and why the answer matters, then ask the smallest question that distinguishes the viable outcomes. ## Recognize Task-Shaped Triggers Use this skill when a requirement is ambiguous, a specification is missing, or the task leaves a material “which approach?” or “should I use X or Y?” choice. It also applies when research has narrowed a decision but taste, priority, private facts, credentials, or authorization still control the outcome. If the job is to verify a claim before presenting it, use **thinking-is-not-knowing**. If proportionate research still leaves no grounded answer, use **i-dont-know-is-the-answer**. This skill owns the middle case: evidence can ground the options and recommendation, but a decision may still need to be carried to the user. ## Keep Research Proportionate Research cost should be proportionate to consequence, uncertainty, and reversibility. Stop researching when a targeted answer from the user is materially cheaper than further investigation. Before asking, state what was checked and why the answer matters so the user can respond without reconstructing the investigation. Course-correct when evidence changes the premise. Sunk cost is no evidence for continuing a wrong course: preserve the objective, explain what changed, and recommend the better direction. ## Preserve Authorization Boundaries Research never implies approval, consent, credentials, or permission for destructive or materially external actions. Discovering how to perform an action does not authorize it. Proceed only within authority already granted; otherwise carry the researched decision and ask for the missing authority.