Skip to content

Arnie016/codex-prompt-templates / skills/model-cost-router

model-cost-router

Verified

Decide when to use cheaper, local, or frontier models and how to measure cost tradeoffs. Use for token limits, model routing, cost-conscious Codex work, DeepSeek/Claude/OpenAI comparisons, prompt caching, or agent workflow economics. Skip when cost routing is enforced by a production gateway, budget policy, fallback policy, or live alerting; use `$auto-skill-build-agent-reliability-loop`. Avoid unverifiable savings claims.

Model Cost Router

Optimize for total successful-task cost, not the cheapest single call.

Routing Heuristic

WorkModel tier
Search, inventory, mechanical editscheap/fast model or local tools
Design decisions, risky code, securitystronger model
Final review of broad changesstronger model
Long-running monitoringno model; use process/log checks
Repeated repo factsmarkdown memory, not model tokens

Measurement Loop

  1. Define the task and quality bar.
  2. Run cheap path once.
  3. Run strong review on the output.
  4. Track elapsed time, fixes needed, and estimated tokens.
  5. Keep routing only if quality and reliability are acceptable.

Guardrails

  • Treat "17x cheaper" or similar claims as marketing until measured locally.
  • Do not route secrets or proprietary code through unknown hosted tools.
  • Prefer caching briefs and memory over re-sending context.
  • Escalate model quality when failure would be more expensive than tokens.
  • For production agents, bind model routing to gateway budgets, cache rules, trace metadata, and eval score gates; route to $auto-skill-build-agent-reliability-loop when cost decisions affect live users or tools.

Pair with $agent-observability to keep a local cost ledger.

Validation

  • Cost claims are measured locally or marked unverified.
  • The cheaper path is kept only if quality and reliability meet the task bar.
  • Production budget/fallback policy is routed to $auto-skill-build-agent-reliability-loop.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 470kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall