Metric map
Ask for p50/p95/p99 latency, error rate, CPU, memory, queue depth, cache hit rate, DB query time, bundle size, and third-party timing before changing code.
🤖 Browser-side performance diagnosis prompt builder
Performance Profiling Prompt Generator
By Omellody Editorial Team · Published · UpdatedUse this existing Omellody prompt utility to turn a slow endpoint, frontend interaction, worker job, or database path into an evidence-first profiling plan with metrics, hypotheses, safe checks, and rollback criteria. The builder runs locally in your browser and does not send project details to Omellody.
Interactive prompt builder
Replace the examples with sanitized project details. The generated prompt updates locally in your browser.
Act as a senior engineer and careful reviewer. Help me create a performance profiling plan from the real context below.
Project context: {context}
Stack or tools: {stack}
Goal: {goal}
Constraints: {constraints}
Return these sections:
1. Direct recommendation with assumptions called out.
2. Inputs needed before acting and questions that block a reliable answer.
3. Decision table: option or hypothesis, evidence needed, risk, next action.
4. Step-by-step workflow with safe checks first.
5. Security, privacy, reliability, and maintainability review.
6. Tests, examples, or validation cases before merge or publication.
7. Rollback, escalation, and owner handoff plan.
8. Final checklist.
Rules: use only supplied facts, mark unknowns, avoid secrets or private data, do not invent undocumented behavior, and explain how to verify the result.
Copy-ready base prompt
Act as a senior engineer and careful reviewer. Help me create a performance profiling plan from the real context below.
Project context: {context}
Stack or tools: {stack}
Goal: {goal}
Constraints: {constraints}
Return these sections:
1. Direct recommendation with assumptions called out.
2. Inputs needed before acting and questions that block a reliable answer.
3. Decision table: option or hypothesis, evidence needed, risk, next action.
4. Step-by-step workflow with safe checks first.
5. Security, privacy, reliability, and maintainability review.
6. Tests, examples, or validation cases before merge or publication.
7. Rollback, escalation, and owner handoff plan.
8. Final checklist.
Rules: use only supplied facts, mark unknowns, avoid secrets or private data, do not invent undocumented behavior, and explain how to verify the result.
Prompt formula and variables
Formula: Symptom + affected path + baseline metric + recent change + observability sources + safety limits + verification gate.
| Variable | What to enter |
|---|---|
{context} | Add specific, sanitized context details. If unknown, write the assumption explicitly. |
{stack} | Add specific, sanitized stack details. If unknown, write the assumption explicitly. |
{goal} | Add specific, sanitized goal details. If unknown, write the assumption explicitly. |
{constraints} | Add specific, sanitized constraints details. If unknown, write the assumption explicitly. |
Hypothesis ladder
Require hypotheses to be ranked by evidence, blast radius, and reversibility instead of asking for random optimization tips.
Safe experiment plan
Demand staging reproduction, feature-flag rollback, before/after dashboards, and owner handoff before production rollout.
Output review table
| Check | Pass condition | Fix if weak |
|---|---|---|
| Grounding | The answer distinguishes facts, assumptions, and unknowns. | Add source snippets, example inputs, or explicit non-goals. |
| Safety | No secrets, customer data, account identifiers, or destructive commands are requested. | Sanitize examples and ask for read-only checks first. |
| Actionability | The plan includes concrete steps, owners, tests, and rollback or review criteria. | Request a checklist with evidence and stop conditions. |
| Verification | The output can be tested with unit cases, logs, traces, fixtures, or source-of-truth docs. | Add expected results and failure examples before using it. |
Source snapshot
| Item | Snapshot |
|---|---|
| Page type | Existing Omellody coding prompt utility refreshed in Red Mode for depth and internal discovery. |
| Demand signal | URL inventory on 2026-05-22 flagged this prompt-family page as thin with limited internal-link depth while traffic radar continued to show AI prompt generator demand. |
| Originality | Omellody-created formula, builder, review criteria, FAQ, source snapshot, and safety guidance. No external repository content copied. |
| Last reviewed | 2026-05-22 |
Related coding prompt tools
FAQ
What should a performance profiling prompt include?
Include the slow path, baseline and current metrics, recent changes, stack, logs or traces available, user impact, constraints, and desired output format.
Should I ask AI for optimization code immediately?
No. First ask for a profiling plan and evidence checklist. Code changes should follow confirmed bottlenecks and tests.
What metrics matter most?
Latency percentiles, error rate, throughput, CPU, memory, database time, cache hit rate, frontend web vitals, and external dependency timing.
Can AI replace profiling tools?
No. AI can structure the investigation, but you still need real traces, logs, flamegraphs, database plans, and load tests.