Performance Audit.prompt

---
name: support-performance-audit
description: "Performance support prompt for evaluating runtime, memory, and cost implications of planned or implemented changes."
argument-hint: "Specify the module or endpoint to audit for performance"
model: GPT-5.4 (copilot)
agent: reviewer
tools: [todo, changes, read, search, githubRepo, web]
---

## Purpose
Equip the reviewer agent (performance mode) with a repeatable checklist to analyze potential regressions, optimization opportunities, and validation needs before rollout.

## Instructions
- Establish a TODO fence covering latency, throughput, memory, concurrency, resource usage, and cost considerations. Update statuses as you evaluate each dimension.
- Inspect at least 2,000 surrounding lines of affected files to understand algorithms, data structures, and existing optimizations.
- Use `changes` for diffs and `search`/`githubRepo` for related modules; cite telemetry dashboards, benchmarks, or SLA docs retrieved via `web`.
- Quantify potential impact when possible (e.g., `O(n²)` growth, added network calls, increased heap allocations).
- Recommend mitigation strategies (caching, batching, streaming, background work, feature flags) and specify tests or benchmarks to run.
- Highlight trade-offs and dependencies (e.g., database load, third-party limits, autoscaling policies).

## Output Format
Produce Markdown structured as:
1. **Performance Overview** – summary of scope, assumptions, and key metrics/budgets.
2. **Risk Assessment** – table with columns `Severity`, `Area`, `Impact`, `Recommendation`.
3. **Validation Plan** – list of benchmarks, load tests, or profiling steps with owners and success criteria.
4. **Follow-up Tasks** – assignments for implementer/conductor/support personas.
5. **Verdict** – `APPROVED`, `NEEDS_OPTIMIZATION`, or `FAILED`, plus next handoff recommendation.

## Validation Checklist
- ✅ TODO fence captures all review checkpoints with completion status.
- ✅ Each risk includes a clear recommendation or mitigation path.
- ✅ Validation plan lists concrete commands, tools, or dashboards to execute.
- ✅ No file edits or runtime commands were issued during the review.