Browse all skills & agents

load-test-plan-designer

Designs a load-test plan from a service's SLOs and endpoint inventory - maps each SLO to load scenarios, defines ramp / soak / spike profiles, sets pass/fail threshold expressions, and outputs a tool-agnostic plan ready to implement in k6 or Gatling. Use when planning a performance test before writing the script; not when choosing the load tool (see load-test-tool-selector in qa-load-testing) or bisecting a perf regression (see perf-regression-bisector).

Modelsonnet

Preloaded skills

slo-load-test-plan

Tools

Read, Grep, Glob

View source

Turns a service's SLO targets and endpoint inventory into a structured, tool-agnostic load-test plan ready to hand to a k6 or Gatling implementer.

When invoked

The agent expects three inputs:

Input	Description	Example
SLO targets / doc	Latency and availability SLOs per endpoint or service tier	"p95 < 300 ms, 99.9% availability"
Endpoint inventory	List of endpoints + expected request weight / relative traffic mix	`GET /api/orders` 60%, `POST /api/checkout` 20%
Expected traffic shape	Peak RPS, daily growth pattern, known spike events	"300 RPS peak, 3x spike on sale days"

Without all three, the agent requests the missing information before producing the plan.

Steps

Read the inputs. Use Read / Grep / Glob over the SLO document, service definitions, and route inventory to collect every SLO, its window, and the traffic weight of each endpoint it governs.
Design the plan. Apply slo-load-test-plan to those inputs: it owns the scenario derivation, the six load profiles, the open versus closed injection model, the threshold expressions traced back to each SLO, the error-budget sizing that calibrates the soak, and the output format.
Emit one Markdown plan, runner-agnostic, with every assumed traffic number listed under Open questions.

Hand-off targets

Choose the tool - ../../qa-load-testing/agents/load-test-tool-selector.md reads the plan's rate profile, duration, and CI gating requirement and recommends k6, JMeter, Gatling, or Locust.
Implement in k6 - preload k6-load-testing.
Implement in Gatling - preload gatling-load-testing.
Gate a CI pipeline on the plan's thresholds - perf-budget-gate.
Bisect a regression once the test is running - ../../qa-load-testing/agents/perf-regression-bisector.md.