deliberatelisted

# Deliberation Loop Three-agent sequential loop that converges on a single recommendation. Eliminates the menu-of-options pattern. Built because Claude's default response to ambiguity is to throw choices back at the user — and some users ("i aint picking anything") explicitly delegate the synthesis. This skill operationalizes the delegation. ## When to fire **FIRE the loop when ANY of:** - About to ask "should I do A or B?" where A and B are both plausible substantive paths - The user has said any of: "you pick", "you decide", "i aint picking", "stop asking me to choose", "just do it", "converge for me", "synthesize and tell me" - Decision involves competing pressures (craft vs. commercial, wound vs. scope, short-term vs. long-term, project A vs. project B, sunk-cost vs. pivot) - Multiple equally-defensible paths exist - The cost of getting it wrong is high enough that single-shot intuition is risky **DO NOT fire the loop for:** - Single-fact lookups ("what version of X are we on?") - Mechanical tasks (file ops, format conversions, renames, lints) - Decisions where one path is clearly correct (e.g., obvious bug fix, unambiguous user request) - Pure information requests - The user explicitly asks for a menu rather than a recommendation ## The three roles ### Agent 1 — Opinion-Drafter Reads the question + full context. Forms an honest opinion. Picks a path or proposes a better one. Gives reasoning. **Names the cost.** **Names what's uncertain.** ~250 words. Agent 1's