propose

name	propose
description	Converges to a recommended approach — the options, their tradeoffs against stated criteria, a pick, and the rationale (plus what would change the call). Use to decide among approaches, or /propose. Not for generating options (brainstorm) or deep-diving one (ideate).

propose

The convergence skill. Given a set of options (from brainstorm/ideate, or stated), it weighs them against explicit criteria, recommends one, and says why — including what evidence would change the call. It closes a decision; it does not open the space (that is brainstorm) or develop a single concept in depth (that is ideate).

It is read-only and advisory: it writes a PROPOSAL.md, never product code. The output is judged by Review — are the criteria explicit and applied evenly, is the recommendation actually supported by the tradeoffs, and is the "what would change my mind" honest?

What this does / does not do

Does: state the decision + the options in play; make the decision criteria explicit (weighted if it matters); compare each option against them evenly; recommend one with the rationale; name the runner-up and the condition under which it wins; state what evidence would flip the call; write PROPOSAL.md + a receipt. Does not: generate the options (that is brainstorm), develop one option in depth (that is ideate), build a proof-of-concept (that is prototype), or write product code.

When NOT to use

• There are no options yet to choose between — use brainstorm to generate them.
• One direction just needs developing, not deciding — use ideate.
• The deciding factor is an untested feasibility unknown — use prototype to get the evidence first, then propose.
• The decision is already made and needs specifying — go to specify.

Inputs

• The decision to make + the candidate options (from brainstorm/ideate, or given).
• The criteria that matter (or elicit them) + any hard constraints.

How to do it

1. Frame the decision. One line: what is being chosen and why it matters now. List the options in play.
2. Make criteria explicit. The factors the decision turns on (cost, time, risk, fit, reversibility, …) — weighted if some matter more. Criteria come before the comparison, not after (no post-hoc justification).
3. Compare evenly. Score each option against each criterion; apply the same lens to all — no thumb on the scale for a favourite.
4. Recommend one. State the pick and the rationale that the comparison supports. If the comparison does not support a clear pick, say so rather than forcing one.
5. Name the runner-up + the flip condition. Which option is second, and the specific condition under which it would win — decisions are contingent, not absolute.
6. State what would change the call. The evidence or change that would flip the recommendation.
7. Write PROPOSAL.md — decision + options + criteria + comparison + recommendation + flip condition.

Output

• PROPOSAL.md at the product root (or proposals/PROPOSAL.md if that dir exists) — persistence: Ephemeral. Decision · options · explicit criteria (weighted) · even comparison · recommendation + rationale · runner-up + flip condition · what would change the call.
• receipt (<workspace>/receipts/propose-receipt-<ts>.json): base schema + decision, options, criteria, recommendation, runner_up, flip_condition, proposal_path. persistence: Ephemeral.

Completion Criteria

Done when: PROPOSAL.md states the decision, lists the options, applies explicit criteria evenly, recommends one with supported rationale, names the runner-up + flip condition, and states what would change the call; a receipt records the decision + recommendation.

Not done if: criteria were written after the pick to justify it (post-hoc); options were compared unevenly (a favourite got a lighter lens); or a clear recommendation was forced where the comparison did not support one.

Next step: prototype gets evidence for an unresolved feasibility factor; specify turns the chosen approach into a task card.

Related Skills

• brainstorm / ideate: produce the options propose decides among (breadth / depth). propose converges; they open.
• prototype: answers a feasibility unknown that the decision hinges on, before proposing.
• adversary: challenges a high-stakes recommendation. Use it against a PROPOSAL.md when the call is high-impact and low-confidence.