func (it QueryIntent) IsSQLIntent() bool

	// scan. ValidateRegistry asserts it at boot and ExecuteIntent re guards it on every execution.
	{
		Name:        IntentVendorPerformanceScorecard,
		Description: "For each staffing vendor and subcontractor the caller's org has recorded, the deterministic performance scorecard: the fill rate (placements over submissions), the submit to interview rate, the average days to submit, and a weighted PerformanceScore, plus the average frozen readiness band of the candidates the vendor submitted BUT only when the vendor has at least ten evaluated submissions (a small cell floor, else the quality is withheld). An outcome aggregate over the vendor's submissions, explicitly NOT a per candidate match score and NOT an employment decision number; the frozen engine stays the only per candidate scorer and this never recomputes a band.",
		SQL:         vendorPerformanceScorecardSQL,
		Params:      nil,
	},
	// vendor_quality_distribution is a BENCH intent (person derived headcount partition, k=10 small cell
	// suppressed): per vendor, the headcount of the vendor's submitted candidates in each frozen readiness
	// band (green, yellow, red) the engine ALREADY assigned. It is registered here so ValidateRegistry
	// asserts it at boot and so ExecuteIntent re guards its SQL, but it NEVER runs through the raw
	// ExecuteIntent path (which hard rejects a bench intent): it is a benchIntent, so the AnswerGraph bench
	// fork routes it to the suppressing VendorQualityDistribution executor, which seeds the complete band
	// partition per vendor with true zeros and suppresses each vendor's cells below k so a sub k vendor band
	// is never a raw count. It reads the frozen candidate_evaluations band on a join, never recomputing.
	{
		Name:        IntentVendorQualityDistribution,
		Description: "For each staffing vendor the caller's org has recorded, the headcount of the vendor's submitted candidates in each frozen readiness band (green, yellow, red) the engine ALREADY assigned, a suppressed per vendor per band distribution (each vendor's bands sum to the vendor's evaluated candidate count), RLS scoped and piped through small cell suppression so a sub k vendor band never surfaces a raw count. Explicitly NOT a match score (the per candidate score and band remain the frozen engine); this reports the band already decided, ordered by vendor name then band.",

vendor_quality_distribution is a BENCH intent (person derived headcount partition, k=10 small cell