More screenshots
Works with the AI tools you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsOpenRouter Model & Cost Optimizer
Design evidence-labelled OpenRouter routing policies, cost projections, and fallback rules for AI workloads.
$8.99
OpenRouter Model & Cost Optimizer
Example session with this skill installed
We use OpenRouter for three tasks and want a routing policy. Priorities: support quality is more important than cost; classification must be fast and cheap; coding changes must be reliable.
Our approximate monthly workload is 50,000 classifications, 8,000 support replies, and 1,200 coding tasks.
Here is a fictional demonstration snapshot supplied by us, not a current market claim:
Economy: input $0.20 / 1M tokens; output $0.80 / 1M tokens.
Balanced: input $1.50 / 1M tokens; output $6.00 / 1M tokens.
Premium: input $8.00 / 1M tokens; output $32.00 / 1M tokens.
Average classification: 400 input and 30 output tokens.
Average support reply: 1,300 input and 280 output tokens.
Average coding task: 4,500 input and 1,200 output tokens.
We need a primary model, fallback, escalation rule, budget controls, and a test plan. Do not modify our routing or use any API key.
- Read your context and instructions
- Compiled the openrouter model
Decision Summary
Using your fictional price snapshot only, the recommended policy is: Economy for classification, Balanced for support replies, and Premium for coding tasks. This matches the priorities: classification stays fast and cheap, support avoids underpowered replies, and coding favors reliability over cost.
No browsing, API key use, or routing modification was performed.
Evidence Ledger
| Type | Item | Source or status | Effect on decision |
|---|---|---|---|
| User-supplied | Workload volumes | 50,000 classifications, 8,000 support replies, 1,200 coding tasks monthly | Used for monthly cost estimates |
| User-supplied | Token averages | Classification 400/30, support 1,300/280, coding 4,500/1,200 input/output | Used for cost estimates |
| User-supplied | Fictional model-tier prices | Economy, Balanced, Premium snapshot | Provisional demo pricing only |
| Recommendation | Tier routing | Policy choice | Needs validation before production |
Cost Formula
runs per month × (input tokens / 1,000,000 × input price
+ output tokens / 1,000,000 × output price)
Recommended Monthly Base Cost
| Task | Route | Monthly cost |
|---|---|---|
| Classification | Economy | $5.20 |
| Support replies | Balanced | $29.04 |
| Coding tasks | Premium | $89.28 |
| Total | Mixed policy | $123.52 |
For comparison: all-Economy would be $11.30, all-Balanced would be $84.78, and all-Premium would be $452.16. The mixed route spends where quality matters without wasting Premium capacity on simple classification.
Routing Policy
| Task | Primary | Fallback | Escalation rules |
|---|---|---|---|
| Classification | Economy | Balanced | Escalate when confidence is below Proposed 0.80, input is out-of-domain, schema validation fails, or result conflicts with deterministic rules |
| Support replies | Balanced | Premium | Escalate for angry customers, billing/account-risk cases, policy-sensitive answers, repeated user dissatisfaction, or answer-quality score below Proposed 4/5 |
| Coding tasks | Premium | Alternate Premium, then human review | Escalate/retry when tests fail, patch touches security/auth/data migration, diff is large, or model cannot explain the change coherently |
Budget Controls
Base budget from supplied numbers is $123.52/month before retries and escalations.
Recommended controls
| Control | Policy |
|---|---|
| Monthly alert | Proposed $150 soft alert |
| Monthly review gate | Proposed $185 investigate routing, retries, and token growth |
| Classification retry | Proposed 1 Balanced retry only after validation failure |
| Support escalation cap | Track Premium escalation rate; investigate if above Proposed 15% |
| Coding cost guard | Require human approval before Premium retries above Proposed 1 retry per task |
| Token drift | Alert if average tokens grow above Proposed 25% from current averages |
| Outage behavior | Use same-tier alternate first; if unavailable, move up one tier for support/coding and degrade classification to queue/retry |
Validation Plan
| Task | Test set | Pass condition |
|---|---|---|
| Classification | Proposed 300 labeled cases, including Proposed 50 borderline/ambiguous cases | Economy remains primary if accuracy is at least Proposed 95% overall and at least Proposed 90% on borderline cases |
| Support | Proposed 120 real or realistic tickets across routine, angry, refund, technical, and policy-sensitive cases | Balanced remains primary if human reviewers rate at least Proposed 90% as usable without major rewrite |
| Coding | Proposed 60 tasks across bug fixes, refactors, tests, and small features | Premium remains primary if tests pass on at least Proposed 90% and reviewer accepts at least Proposed 85% without material redesign |
| Fallback drill | Force simulated failure for each primary route | System must route to fallback correctly in Proposed 100% of drill cases |
Open Questions
Current OpenRouter model names, live pricing, latency, context limits, and provider availability were not verified. The smallest next step would be to collect a dated OpenRouter snapshot and run an offline validation batch with logged token counts, outputs, reviewer scores, latency, retry rate, and escalation rate.
Connects securely to your tools. The creator never sees your data.
What you get
About this skill
The problem
Managing costs and reliability across multiple LLM providers is complex. Developers often overpay for simple tasks or suffer outages because they lack a systematic model routing and fallback strategy.
What it does
- Maps your specific workload into task classes based on risk, volume, and latency needs.
- Defines primary and fallback model routes for every task type to ensure high availability.
- Establishes observable escalation triggers and validation gates for quality control.
- Calculates evidence-backed cost projections using your specific token usage and pricing data.
Frameworks & tools
OpenRouter API, LLM orchestration frameworks, and JSON-based routing configurations.
Why this beats prompting it yourself
Standard prompts often ignore the nuance of cost-to-quality ratios or fail to provide measurable validation gates. This skill enforces a rigorous evidence-based approach, marking every non-verified number as Proposed to prevent accidental reliance on hallucinated benchmarks or outdated pricing.
Use cases
- Optimizing OpenRouter spend by routing routine tasks to cheaper, specialized models.
- Architecting multi-model redundancy to handle provider outages or rate limits.
- Setting up automated quality gates that trigger escalation to high-reasoning models.
- Projecting monthly API budgets for new AI features before deployment.
Known limitations
Does not provide real-time pricing from memory; requires user-supplied dated snapshots or permission to verify sources. Does not store or manage API keys.
How to install
Works the same in every agent - Claude, Cursor, Codex, Copilot and 20+ more.
- 1
Download the ZIP
Free skills download straight away. Paid skills unlock right after purchase.
- 2
Unzip into your skills folder
Every agent reads skills from one folder on your machine. Drop the unzipped folder in there.
- 3
Ask your agent to use it
Restart the agent if it was already running. It picks the skill up automatically - no config needed.
Skills folder by agent
Click the path to copy it. Create the folder if it does not exist yet.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean 16 days ago
- Passed all security checks, Safe to install