Terminal
Terminal/Skill
Reviews Amazon Bedrock workloads against the Well-Architected Framework for security, performance, quotas, cost and resilience, with a shareable report.
By Amazon Web Services
Works with
npx skills add aws/tools-for-devops-agent --skill bedrock-operation-reviewConduct a comprehensive operational review of Amazon Bedrock workloads aligned with the AWS Well-Architected Framework and Amazon Bedrock best practices.
This skill uses the AWS Bedrock, Bedrock Agent, CloudWatch, Service Quotas, and EC2 APIs only — all data is collected through native AWS control-plane APIs and CloudWatch metrics. It performs no data-plane model invocations and reads no prompt or response content.
Activate this skill when the user asks to:
Ask the user which accounts and regions to review. Accept:
If no scope is given, default to all configured account regions and all pillars. Default the analysis window to the last 7 days unless the user specifies a range (historical windows older than ~2 weeks may have reduced CloudWatch resolution).
Per account/region, begin by checking whether the account actually uses Bedrock in
this region. Do not use bedrock.ListFoundationModels as an activity signal — it
returns the regional model catalog available to the account and is generally non-empty
in every supported region regardless of usage, so it is treated as catalog data only
(see ListFoundationModels).
Instead, determine activity from account-owned resources and AWS/Bedrock metrics:
bedrockagent.ListAgents, bedrockagent.ListKnowledgeBases,
bedrock.ListGuardrails, bedrock.ListCustomModels,
bedrock.ListProvisionedModelThroughputs, bedrock.ListInferenceProfiles
(application-scoped), and bedrock.ListModelCustomizationJobs.cloudwatch.ListMetrics for the AWS/Bedrock namespace.If no account-owned Bedrock resources exist AND CloudWatch shows no AWS/Bedrock
metrics for the region, record "No Bedrock activity detected — skipping pillar
analysis" for that region and move on. Do not generate empty findings tables for
inactive regions.
For regions where Bedrock is active, collect the full resource inventory:
bedrock.ListFoundationModels # catalog reference only: model
# availability + lifecycle status
# (NOT an account-activity signal)
bedrock.ListGuardrails / GetGuardrail # configured guardrails
bedrock.GetModelInvocationLoggingConfiguration
bedrock.ListInferenceProfiles / GetInferenceProfile
bedrock.ListPromptRouters / GetPromptRouter
bedrock.ListProvisionedModelThroughputs / GetProvisionedModelThroughput
bedrock.ListCustomModels / GetCustomModel
bedrock.ListModelCustomizationJobs / GetModelCustomizationJob
bedrockagent.ListAgents / GetAgent # agents (draft version)
bedrockagent.ListAgentVersions / GetAgentVersion # deployed versions — read only the
# per-version model ID, orchestration
# type, and whether a prompt override is
# present. Do NOT read the override
# base-prompt template text.
bedrockagent.ListAgentAliases
bedrockagent.ListKnowledgeBases / GetKnowledgeBase
bedrockagent.ListDataSources / GetDataSource
bedrockagent.ListPrompts # Prompt Management: presence/metadata
# only (identifiers, versions, counts).
# Do NOT call GetPrompt — it returns
# prompt template/variant content, which
# this skill does not read.
Capture per resource: identifiers, ARNs, status, creation/update timestamps, encryption configuration (AWS-managed vs customer-managed KMS key), and any VPC configuration.
AWS/Bedrock)Before querying metrics, load the authoritative thresholds reference:
read_skill_resource(skill_id="bedrock-operation-review", path="references/metrics-thresholds.md")
Use the thresholds from that file when classifying metric values as Normal, Warning, or Critical throughout this step and Step 4.
Discover which models are actually invoked with cloudwatch.ListMetrics
(dimension ModelId), then pull metric data with cloudwatch.GetMetricData.
Use the user's selected window; default to 7 days.
Key model-level metrics (dimension ModelId):
| Metric | Stat | Signal |
|---|---|---|
| Invocations | Sum | Volume / denominator for rate calcs |
| InvocationThrottles | Sum | Throttling / quota pressure |
| InvocationServerErrors | Sum | Server-side failures |
| InvocationClientErrors | Sum | Bad input / blocked content |
| InvocationLatency | Average, p95 | End-to-end latency |
| TimeToFirstToken | Average, p95 | Perceived latency (streaming) |
| InputTokenCount | Sum | Input token volume |
| OutputTokenCount | Sum | Output token volume |
| InvocationsIntervened | Sum | Guardrail interventions |
| TextUnitCount | Sum | Guardrail text-unit consumption |
| CacheReadInputTokenCount | Sum | Prompt cache reads |
| CacheWriteInputTokenCount | Sum | Prompt cache writes |
GPU metrics for self-managed workloads live in the CWAgent namespace
(nvidia_smi_utilization_gpu, nvidia_smi_memory_util).
cloudwatch.GetMetricData allows up to 500 metric-data queries per call — batch
requests and paginate when a region has many invoked models.
Before evaluating findings, load the best-practices checklist:
read_skill_resource(skill_id="bedrock-operation-review", path="references/best-practices-checklist.md")
Use the checklist as the canonical list of items to evaluate for each pillar. Mark each item as ✅ Pass, ⚠️ Warning, ❌ Fail, or ➖ Not Applicable, and generate a finding for every Fail or Warning.
Evaluate all collected data across the pillars below and assign a severity to every finding: CRITICAL, HIGH, MEDIUM, LOW, or INFO. The pillars mirror the Bedrock operational review structure: Security, Performance, Service Quotas, Cost Optimization, Resilience.
Ref: Security in Amazon Bedrock
InvocationClientErrors (>5% of Invocations)
indicates problematic content reaching models — filter before invocation → MEDIUM.
InvocationsIntervened <5% of invocations on an active guardrail suggests
under-configuration; >15% suggests policy over-tuning → review.*
without justification → MEDIUM. Follow least privilege for inference endpoints.
A comprehensive least-privilege audit of all IAM policies is out of scope for this
skill — flag obvious *-resource findings only, at INFO severity.
IAM for Amazon BedrockRef: Monitoring Amazon Bedrock
InvocationLatency p95, TimeToFirstToken p95,
throttle rate (InvocationThrottles / Invocations), and server error rate. High
throttle rate → HIGH (add retries with exponential backoff + jitter, request quota
increase, spread load, use CRIS). Prefer smaller models and streaming for
latency-sensitive apps.
Improve Bedrock performanceGetAgentVersion) against the draft. Aliases pinned to outdated versions with
different model IDs or orchestration types than the current draft → MEDIUM (may
miss performance or capability improvements).default, flex, priority,
reserved) matches workload criticality. Latency-sensitive apps on Flex, or batch
workloads on Priority, are misaligned → MEDIUM.
Service tiersRef: Bedrock quotas
servicequotas.GetServiceQuota. Utilization >75% → MEDIUM (request increase before
throttling); sustained near the limit → HIGH.Ref: Bedrock pricing
InvocationLatency p99
3000 ms AND
InputTokenCount>1M/month AND throttle rate >5% → high-priority candidate → MEDIUM opportunity. (us-east-1, us-west-2) Model distillation
CacheReadInputTokenCount / (CacheReadInputTokenCount + CacheWriteInputTokenCount).
Statuses: Not Supported, Not Enabled, Misconfigured (writes but no reads),
Underutilized (<50%), Optimized (≥50%). Input-heavy repeated context that isn't
cached → MEDIUM opportunity (up to ~85% latency and ~90% cost reduction on cached
prefixes).
Prompt cachingProfile InputTokenCount / (Profile InputTokenCount + Base Model InputTokenCount).
Low adoption (traffic bypassing the inference profile by invoking base models
directly) reduces burst resilience and, for global profiles, forgoes ~10% savings →
MEDIUM.Generate a shareable report artifact for the review.
Artifact naming: bedrock-review-<account-id>-<region>-<YYYY-MM-DD>.md
Example: bedrock-review-123456789012-us-east-1-2026-04-29.md
Structure the Markdown document with:
# Amazon Bedrock Operational Review — <account-id> / <region>
Date: <YYYY-MM-DD> | Analysis window: <start> to <end>
Pillars reviewed: <list>
For each of Security, Performance, Service Quotas, Cost Optimization, Resilience:
| # | Finding | Severity | Current State | Recommendation |
| Metric | Model | Stat | Value | Status | Finding |
| Quota | Value | Observed P95 | Utilization % | Risk |
| Opportunity | Signal | Est. Impact | Effort |
| # | Finding | Severity | Pillar | Effort | Impact |
| Severity | Definition | SLA |
|---|---|---|
| CRITICAL | Immediate risk to availability, security, or data integrity | Fix within 24–48 hours |
| HIGH | Significant gap that could lead to incidents | Fix within 1 week |
| MEDIUM | Notable improvement opportunity | Plan within 30 days |
| LOW | Minor optimization or hardening | Address when convenient |
| INFO | Observation, no action required | N/A |
Some checks only run in specific regions:
Skip a check silently in unsupported regions rather than reporting a failure.
cloudwatch.GetMetricData caps at 500 metric-data queries per call — batch and
paginate for accounts with many invoked models.nvidia_smi_* metrics are absent and GPU signals can't be evaluated.This skill collects data exclusively through native AWS APIs
(bedrock, bedrockagent, cloudwatch, servicequotas, ec2). It does not: