fpf-selection

## What this skill IS This is **deliberate choice with trade-off awareness**, not picking the obvious winner. You are mapping the Pareto front, stating your selection policy BEFORE applying it, and preserving stepping stones. **Key discipline:** Never collapse NQD to a single score. Hold the Pareto front. State why you're choosing one point on the front over others. ## Prerequisites - SPORT-* with NQD-characterized variants (≥3) - CHR-* with indicators and comparison rules - PROB-* with acceptance criteria - If comparing variants with measurable indicators: PAR-* parity plan (invoke `/fpf-parity` first) ## Output `.fpf/decisions/SEL-${CLAUDE_SESSION_ID}--<slug>.md` ## Constraints (quality bar) - **C1:** Selection policy stated BEFORE applying — not post-hoc rationalized - **C2:** Dominance table complete — every variant on every CHR indicator (multi-dimensional Q) - **C3:** Pareto front identified (non-dominated variants) - **C4:** ≥1 stepping-stone bet — non-dominant variant preserved with conditions to reconsider - **C5:** "What would change this" section filled — what evidence or constraint shift would reverse selection - **C6:** Missing inputs (portfolio, passport, problem card) flagged explicitly - **C7:** WLNK comparison — dominance analysis MUST consider each variant's weakest link. A variant whose weakest link is weaker than another's is dominated on that axis. - **C8:** MONO check — if the selected variant is more complex than alternatives (more components, more abstractions), state why the added complexity is justified given the new weak links it introduces. ## Format ```markdown # Selection Record - **ID:** SEL-... **Portfolio:** SPORT-... **Passport:** CHR-... - **valid_until:** YYYY-MM-DD ## ConstraintFit eligibility (admission gate) (hard constraints — variants failing these are INELIGIBLE regardless of NQD scores: safety, compliance, cost ceiling, ...) | Variant | Must-constraints | Pass/Fail | ## Selection policy (stated before applying — e.g., "maximize latency within cost constraint" or "learning value first") - Main criterion: ... - Tie-breakers (in order): ... - Minimum trust level: (what counts as "sufficiently verified" for this selection) ## Dominance table | Variant | [CHR indicator 1] | [CHR indicator 2] | ... | WLNK | N | D_p | Dominated by | ## Pareto front (non-dominated variants) ## Decision Selected: V[N]. Rationale. What was sacrificed. Why this point on the Pareto front. ## Stepping-stone bets | Variant | Why preserve | What future space it opens | Condition to reconsider | ## MONO assessment (If selected variant is more complex than alternatives: what new weak links, and why is added complexity justified?) ## What would change this (what evidence, constraint change, or new information would reverse this selection) ## valid_until (when does this selection go stale?) ```