# Fixture-only baseline full context prompt

This synthetic baseline prompt represents the unoptimized full-context side of a future matched benchmark. It is dry-run-only and not a hosted API token/cost savings claim. A real run must replace this prose with sanitized project evidence, a real success_command, provider-measured primary token/cost fields, matched successful tasks, a 10%p failure-rate guardrail, human corrections, and shifted-cost accounting.

Include exact protected evidence such as paths, IDs, stack frames, JSON keys, and artifact receipt fallback handles without semantic rewriting.

Metric checklist for real replacements: tokens_per_successful_task, total_cost_with_shift_usd, external_cost_usd, matched_successful_task evidence, 10%p failure-rate guardrail, and proxy-byte caveat.