| name | bedrock-tier-advisor |
| display_name | Bedrock Tier & Variant Advisor |
| description | Recommends the right Bedrock service tier (Flex/Standard/Priority/Reserved/Batch) and inference variant (Global/Regional) based on workload characteristics. Activate when the user asks which tier to use, wants to optimize cost vs performance, or hasn't specified a tier for a pricing estimate. Reads guidance from bedrock-tier-guidance.md which can be refreshed from AWS documentation. |
| icon | 🎯 |
| trigger | bedrock tier advice |
| depends-on | ["bedrock-pricing"] |
| inputs | [{"name":"workload_type","description":"Type of workload: production, dev-test, batch, mission-critical. Helps narrow tier recommendation.","type":"string","required":false},{"name":"model","description":"Model name (e.g. 'Claude Sonnet 4.6'). Used to discover available tiers from cache.","type":"string","required":false},{"name":"region","description":"AWS region code (e.g. us-east-1). Used to discover available variants.","type":"string","required":false}] |
| tools | ["run_python","file_read"] |
Overview
Advises which Bedrock service tier and cross-region inference variant to use for a given workload. Combines static guidance (from bedrock-tier-guidance.md) with dynamic discovery (from the pricing cache) to produce a recommendation.
Part of the skill family: bedrock-pricing, agentcore-pricing, bedrock-capacity, agent-business-value, bedrock-tier-advisor (this).
Guidance File
The tier and variant guidance is stored in bedrock-tier-guidance.md in the same directory as this skill file.
Workflow
Step 1: Load Guidance
Read bedrock-tier-guidance.md for the decision framework.
Step 2: Discover Available Tiers for the Model
Use extract_bedrock_model_prices(results, all_tiers=True) from the bedrock-pricing script to see which tiers and variants actually exist for the user's model in their region.
import sys, os
sys.argv = ['bedrock_pricing.py']
_p = os.path.join(os.getcwd(), "skills/bedrock-pricing/scripts/bedrock_pricing.py")
if not os.path.exists(_p):
_p = os.path.expanduser("~/.quickwork/skills/bedrock-pricing/scripts/bedrock_pricing.py")
exec(open(_p).read())
home = os.path.expanduser("~/bedrock_cache")
results = query_model_pricing(home, region_filter="us-east-1", model_filter="Claude Sonnet 4.6")
all_prices = extract_bedrock_model_prices(results, all_tiers=True)
Step 3: Apply Decision Framework
Match the user's workload characteristics against the guidance:
| User Says | Recommend |
|---|
| "production agent" / "customer-facing" | Standard Global (with prompt caching if available) |
| "dev/test" / "experimenting" | Flex if available, else Standard |
|
