Problem Framing

A solution accepted as a problem can only be built, never tested. Framing climbs from the ask to a cost a specific person bears, then names the one belief that, if false, collapses the whole thing.

Mode. Standalone frames the supplied idea or complaint with its known audience and constraints and returns the problem statement, unknown facts kept unknown (steps 1 to 5). The lifecycle runs every step, writes 00-frame.md and gate 0. Both per operating modes. Load pressure-testing for the rounds and evidence-ledger when tags need checking. Read framing methods when a step needs its worked detail or examples.

Procedure

1. Capture the ask verbatim, word for word, before interpreting it: the phrasing carries the assumption you are about to extract. A ticket is read with its history and tagged [doc:*]. An anomaly (a watched metric outside its bands, an incident) starts from the metric, baseline, breach and window, tagged [data:*] or [doc:*], and climbs from the observed cost; its riskiest assumption is usually about the cause. Where PRODUCT.md already evidences the problem and ICP, the work may qualify for the feature track (lifecycle-setup). Done when the ask is quoted exactly with its source.
2. Climb the ladder. One "why does that matter" at a time, in conversation. Stop when the statement names a cost a specific person bears and the next why would produce a truism. It passes the "so what" test when a role (not an organization) bears the cost, today's workaround is named, and its cost is named (unquantified is acceptable as [assumption:unvalidated]). Then run the solution-language check (gate 0.1, word list in gate-checks); the product's own nouns are defined in PRODUCT.md → Language and used as defined. Done when the statement passes both checks.

3. Run the context round. Retrieve known ICP, why-now and prior-art facts first (the prior-art search runs before the round where web.search resolved, so that question arrives answered and tagged), then ask what remains as a pressure-testing round. Work each answer until it holds:

   • ICP. One primary, others explicitly secondary, with all five fields: segment, size, trigger, buying power, reachability. Push hardest on size (a number with a source tag; bottom-up beats a borrowed market report) and reachability (five people the person could get on a call this week). When reachability fails, say plainly that phase 1 will stall and that finding reachable users is now the first task.
   • Why now. A dated change: behavior shifted, cost collapsed, a constraint lifted. With none evidenced, say so and label candidate structural reasons as hypotheses; "the technology is good enough now" needs the capability and the line it crossed.
   • Prior art. The two closest existing solutions, each classified solved well and adopted, solved badly but adopted, or solved and not adopted: a shallow scan, with product-analysis (competition) when more is needed. "Solved well, adopted" is said directly: stopping here is a legitimate phase 0 outcome and costs a conversation instead of a quarter.

   Done when the ICP has all five fields, why-now has a dated change or a stated gap, and prior art has two classified entries.

4. Put the riskiest assumption to the person. List the beliefs the frame requires and score each on confidence against collapse. The two or three low-confidence, high-collapse candidates become one round question: each as a falsifiable sentence with the observation that would falsify it and what phase 1 would then test, and a recommendation. An unfalsifiable candidate is a preference: label it and offer the next. Record **Chosen by:** under ## Riskiest assumption with the person's answer (gate 0.5). Done when one chosen assumption has its falsifier.
5. List the open questions: what the frame could not settle (a population nobody has reached, an access rule, a constraint nobody owns), who can answer each, and which phase inherits it. Done when each open question has an owner and a phase.
6. Deliver. Standalone returns the framing and its uncertainties. The lifecycle writes .builderos/initiatives/{initiative}/00-frame.md per the frame template and runs gate 0 per gate-checks; on a pass, present the frame and ask whoever owns the problem to accept it, then record it with their answer (scripts/bos.mjs record 0 --judged ... --accepted-by "who" where commands run). A failed gate gets the refusal. A project with no .builderos/ is reported as uninitialized: initialization is its own step. Done when the frame is delivered and, in the lifecycle, gate 0 is recorded with the person's acceptance.

Watch for

Capabilities

Phase 0 needs no data capability: an idea with nothing connected is the normal entry, and analytics waits for later phases.

Output Contract

## FRAME COMPLETE, followed by the problem, the ICP, the riskiest assumption and who chose it, the open questions, the gate verdict, who accepted, and the research target phase 1 inherits. Standalone output follows the requested format and destination, with no gate claims.