optimize-codex-usagelisted

# Optimize Codex Usage Reduce total Codex consumption while preserving the requested outcome, authorization boundaries, safety checks, and completion evidence. ## Load Policy - For a conceptual explanation or a narrow current-product fact, use this file and current official OpenAI documentation only. Do not inspect a repository or load the audit reference unless the request needs it. - For an audit, design change, or implementation involving Skills, `AGENTS.md`, MCP, context, history, tools, validation, or reporting, read `references/context-audit.md` before changing files. - Add another Axiom skill only when the requested implementation crosses into that skill's distinct owner: an AGENTS instruction system, traceable Git state, or a persistent system change. Usage analysis alone needs only this skill. ## Measurement Boundary - Use exact token, credit, cache, reasoning, or tool metrics only when the host exposes them for the scoped run. Axiom cannot recover hidden usage data. - Otherwise label UTF-8 bytes, words, lines, reference counts, tool calls, and route-chain sizes as proxies or estimates. - Do not claim a percentage saving without a repeatable before/after workload and equivalent quality acceptance. - Fetch current official OpenAI documentation only when a volatile Codex fact affects the decision. Never place current prices, plan limits, or model quotas in always-loaded instructions. ## Non-Negotiable Boundaries - Define the required qual