revfactory Performance Optimizer — performance optimization pipeline
thisof profiling→bottleneckanalysis→optimization→benchmark inthisbefore teamthis to in count.
execution
inthisbefore team — 5peoplethis SendMessageas direct and verification.
inthisbefore setup
| inthisbefore | day | role | type |
|---|---|---|---|
| profiler | .claude/agents/profiler.md |
CPU, memory, I/O, network profiling | general-purpose |
| bottleneck-analyst | .claude/agents/bottleneck-analyst.md |
hotspot identification, , impactcalculate | general-purpose |
| optimization-engineer | .claude/agents/optimization-engineer.md |
code/query/architecture optimization | general-purpose |
| benchmark-manager | .claude/agents/benchmark-manager.md |
test, execution, analysis | general-purpose |
| perf-reviewer | .claude/agents/perf-reviewer.md |
verification, regression, finalreport | general-purpose |
workflow
Phase 1: (this direct count)
- user from :
- optimization upper: code, API, query, system before
- performance target: between, throughput, memory for etc. -based count
- ** stack**: language, framework, DB, infrastructure
- **current ** (optional): -based upper, error log, user only
- existing day (optional): profiling result, execution plan, code
_workspace/project rootin creation- to
_workspace/00_input.mdin - existing daythis
_workspace/in and corresponding Phase cases - request scopein execution decision
Phase 2: team setup and execution
| responsible | of | |||
|---|---|---|---|---|
| 1 | profiling | profiler | _workspace/01_profiling_report.md |
|
| 2 | bottleneck analysis | bottleneck-analyst | 1 | _workspace/02_bottleneck_analysis.md |
| 3 | optimization | optimization-engineer | 2 | _workspace/03_optimization_plan.md |
| 4 | benchmark verification | benchmark-manager | 1, 3 | _workspace/04_benchmark_results.md |
| 5 | performance review | perf-reviewer | 1~4 | _workspace/05_review_report.md |
team between :
- profiler completed → analystto hotspot·resources pattern before, benchmarkto baseline count before
- analyst completed → engineerto bottleneck priority· before, benchmarkto performance budget before
- engineer completed → benchmarkto optimization code· count before
- benchmark completed → reviewerto analysis result before
- reviewer all verification. 🔴 required modification corresponding inthisbeforeto modification request → → verification (maximum 2)
Phase 3: integrated and final
_workspace/within all day confirmation- review reportof 🔴 required modificationthis confirmation
- final userto report:
- profiling —
01_profiling_report.md - bottleneck analysis —
02_bottleneck_analysis.md - optimization plan —
03_optimization_plan.md - benchmark —
04_benchmark_results.md - review report —
05_review_report.md
- profiling —
per
| user request pattern | execution | inthisbefore |
|---|---|---|
| "performance before optimization" | ** pipeline** | 5people before |
| "this code profiling" | **profiling ** | profiler + reviewer |
| "this query optimization" | **query optimization ** | profiler + analyst + engineer + reviewer |
| "benchmark test " | **benchmark ** | benchmark + reviewer |
| "this optimization result verification" | **review ** | reviewer |
data before as
| strategy | foralso | |
|---|---|---|
| day | _workspace/ |
week and shared |
| message | SendMessage | real-time core information before, modification request |
| TaskCreate/TaskUpdate | in progress upper tracking, of |
error
| error type | strategy |
|---|---|
| profiling also execution impossible | code -based analysisas , also configuration guide provided |
| execution | benchmark script + execution guide provided, expected result this |
| performance target | (P95 < 200ms, error < 1%) target proposal |
| inthisbefore failure | 1 retry → failure corresponding this in progress, review reportin people |
| reviewfrom 🔴 | corresponding inthisbeforein modification request → → verification (maximum 2) |
test
normal
****: "Django REST APIof betweenthis P95 criteria 2seconds, 500ms or lessas this . code analysisand optimization." ** result**:
- profiling: Django middleware, ORM query, -ize and analysis, N+1 query detection
- bottleneckanalysis: DB query before betweenof 70% → ORM N+1 + index
- optimization: select_related/prefetch_related -basedfor, index addition, queryset caching
- benchmark: k6 scriptas beforeafter , P95 2s → 400ms verification
- review: confirmation, regression risk evaluation
existing day for
****: "this profiling result analysisand optimization " + flame graph ** result**:
- existing profiling data
_workspace/01_profiling_report.mdas for - profiler cases analyst + engineer + benchmark + reviewer
error
****: "this slow " ** result**:
- code -based analysis profiling count
- day-based performance pattern list -basedfor
- "detailed profiling also -basedfor " people
inthisbeforeper extension
| as | -ize upper inthisbefore | role | |
|---|---|---|---|
| query-optimization-patterns | .claude/skills/query-optimization-patterns/skill.md |
bottleneck-analyst, optimization-engineer | execution plan analysis, index , N+1 resolution, thisthis |
| caching-strategy-selector | .claude/skills/caching-strategy-selector/skill.md |
optimization-engineer | Cache Aside/Write Through/Behind, invalid-ize, |