bsafronov Challenge Plan
Expose what could change the user's decision, then help them resolve or explicitly accept those uncertainties.
- Anchor the critique in the intended outcome, scope, constraints, and decisions already made. Inspect the supplied plan and relevant evidence before asking for context. Distinguish facts to verify from preferences only the user can decide; retrieve available facts yourself. Mark unavailable evidence as unknown instead of turning it into an assumed constraint.
- Prioritize assumptions whose failure would change feasibility, cost, user value, or reversibility. Use a concrete counterexample, failure scenario, or comparison with a credible alternative to test each material claim. Separate evidence-backed problems from hypotheses; a request for criticism does not require inventing objections to a supported choice.
- Ask the smallest useful round of questions, respecting the user's depth and time budget. Give your recommended answer and its decisive trade-off when evidence supports one; otherwise name the fact or preference it depends on. Ask prerequisite decisions first. Group only questions that can be answered independently, and wait for answers before developing dependent branches.
- Update the critique after each answer or new fact. Retire resolved objections, retain accepted trade-offs, and focus on what remains consequential. Reopen a decision only when new evidence changes its basis. If uncertainty is empirical, prefer a bounded experiment with an observable success condition over more speculative questioning.
- End with the supported choices, material open questions or accepted risks, and the next decision or evidence that would change the conclusion. A well-supported plan may need no change. Stop when the consequential issues in the requested scope are resolved or explicitly recorded, or the user asks to stop; do not require every imaginable branch to be explored.
The user owns the decision. Critique alone authorizes no implementation or external action. Follow any separately authorized next task without requiring another confirmation merely to close this discussion. Specification writing, domain modeling, and independent defect review remain separate outcomes when requested; none is a prerequisite for this Skill.