Agent Skills › victorgarciaesgi/regle › ts-memory-baseline

ts-memory-baseline

GitHub

定义 TypeScript 内存使用基准测试流程，用于测量编译器诊断和类型检查的内存消耗。通过对比新旧基线中位数，评估泛型变更或性能回归，确保构建和测试稳定性。

.cursor/skills/ts-memory-baseline/SKILL.md victorgarciaesgi/regle

Trigger Scenarios

检查 TS 服务器压力检测类型级性能回归重型泛型类型变更前后的对比

Install

npx skills add victorgarciaesgi/regle --skill ts-memory-baseline -g -y

More Options

Use without installing

npx skills use victorgarciaesgi/regle@ts-memory-baseline

指定 Agent (Claude Code)

npx skills add victorgarciaesgi/regle --skill ts-memory-baseline -a claude-code -g -y

安装 repo 全部 skill

npx skills add victorgarciaesgi/regle --all -g -y

预览 repo 内 skill

npx skills add victorgarciaesgi/regle --list

SKILL.md

Frontmatter

{
    "name": "ts-memory-baseline",
    "metadata": {
        "author": "regle",
        "version": "2026.03.19"
    },
    "description": "Measure and compare TypeScript memory usage for @regle\/core and test typechecks. Use when checking TS server pressure, type-level perf regressions, or before\/after heavy generic typing changes."
}

TypeScript memory baseline procedure

This skill defines a reproducible process to measure TypeScript memory usage in this repository.

Hard requirements

Use Node 24 for all measurements.
Run pnpm build before any test-related measurement.
pnpm run test:typecheck should never fail. If it fails, treat the run as invalid for regression tracking.

Environment setup

PATH="/Users/victorgarcia/.nvm/versions/node/v24.15.0/bin:$PATH"
node -v
pnpm -v

Expected:

Node v24.x
pnpm 11.20.0 (or project-compatible version)

Measurement runbook (must follow)

Ensure no extra monorepo tasks are running.
Use the exact Node 24 PATH prefix in every command.
Run one warm-up pass that is discarded.
Run 3 measured passes and compare medians.
Save only the final medians in tests/TS_MEMORY_BASELINE_RESULTS.md.

Measurement commands

A. @regle/core compiler diagnostics

PATH="/Users/victorgarcia/.nvm/versions/node/v24.15.0/bin:$PATH" \
./node_modules/.bin/tsc -p "packages/core/tsconfig.json" --noEmit --extendedDiagnostics

Capture:

Memory used
Instantiations
Check time

B. Full type tests memory

Run build first:

PATH="/Users/victorgarcia/.nvm/versions/node/v24.15.0/bin:$PATH" pnpm build

Then:

PATH="/Users/victorgarcia/.nvm/versions/node/v24.15.0/bin:$PATH" \
/usr/bin/time -l pnpm run test:typecheck

Capture:

exit code (must be 0)
maximum resident set size
peak memory footprint
real/user/sys

Recommended sequence:

# warm-up (discard)
PATH="/Users/victorgarcia/.nvm/versions/node/v24.15.0/bin:$PATH" \
/usr/bin/time -l corepack pnpm run test:typecheck

# measured runs (collect all 3)
PATH="/Users/victorgarcia/.nvm/versions/node/v24.15.0/bin:$PATH" \
/usr/bin/time -l corepack pnpm run test:typecheck
PATH="/Users/victorgarcia/.nvm/versions/node/v24.15.0/bin:$PATH" \
/usr/bin/time -l corepack pnpm run test:typecheck
PATH="/Users/victorgarcia/.nvm/versions/node/v24.15.0/bin:$PATH" \
/usr/bin/time -l corepack pnpm run test:typecheck

Comparison protocol

Use median deltas because absolute values can vary with machine/CPU load.

Compute medians for old baseline and new run.
Compute delta % = (new - old) / old * 100.
Evaluate with these thresholds:
- Likely noise: within ±5% for memory metrics.
- Watch zone: >5% and <=10% increase.
- Regression candidate: >10% increase.
For Check time and real time, treat as noisy:
- investigate only if increase is >15% and reproducible across runs.
If pnpm run test:typecheck exit code is non-zero, invalidate the run (do not update baseline).