AgentSkillsCN

Cost Analysis

结合优化建议，深入分析基础设施与运营成本。

SKILL.md
--- frontmatter
name: Cost Analysis
description: Analyze infrastructure and operational costs with optimization recommendations
category: operations
version: 1.0.0
triggers:
  - monthly-review
  - cost-anomaly
  - budget-threshold
globs: "**/infrastructure/**,**/*.tf"

Cost Analysis Skill

Analyze infrastructure and operational costs with optimization recommendations.

Trigger Conditions

  • •Monthly/quarterly cost review cycle
  • •Budget threshold breach alert
  • •New resource provisioning request
  • •User invokes with "cost analysis" or "optimize spending"

Input Contract

  • •Required: Resource inventory or billing data
  • •Required: Time period for analysis
  • •Optional: Budget targets, comparison baselines

Output Contract

  • •Cost breakdown by service, team, and resource type
  • •Cost-per-transaction trending
  • •Optimization recommendations with savings estimates
  • •Reserved vs. on-demand vs. spot mix recommendation

Tool Permissions

  • •Read: Billing data, resource configs, usage metrics
  • •Write: Cost analysis reports
  • •Search: Resource tags and attribution data

Execution Steps

  1. •Collect billing data for the analysis period
  2. •Attribute costs by service and team using resource tags
  3. •Identify unattributed costs
  4. •Calculate cost-per-transaction trends
  5. •Identify top cost drivers and anomalies
  6. •Generate optimization recommendations (rightsizing, reservations, spot)
  7. •Estimate savings per recommendation

Success Criteria

  • •100% of costs attributed to service/team
  • •Cost anomalies flagged with explanation
  • •Optimization recommendations include estimated savings
  • •Cost-per-transaction trend reported

Escalation Rules

  • •Escalate if daily spend exceeds baseline by >20%
  • •Escalate if unattributed costs exceed 10% of total
  • •Escalate if optimization requires architectural changes

Example Invocations

Input: "Analyze our AWS costs for January and find savings"

Output: Total: $45,200. Top 3: RDS ($18K — oversized instances, recommend r6g.xlarge→r6g.large, save $4K/mo), EKS ($12K — 40% idle capacity, recommend cluster autoscaler tuning), S3 ($8K — no lifecycle policy, archive to Glacier after 30d, save $3K/mo). Total potential savings: $7K/month.