菜单
Execute Python code locally with marketplace API access for 90%+ token savings on bulk operations. Activates when user requests bulk operations (10+ files), complex multi-step workflows, iterative processing, or mentions efficiency/performance.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/muhammedadnank/Antigravity-Skills --skill code-execution命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
This skill should be used for advanced LLM evaluation: LLM-as-judge systems, direct scoring, pairwise comparison, rubric calibration, evaluator bias mitigation, confidence scoring, and automated quality assessment.
Automated end-to-end UI testing and verification on an Android Emulator using ADB.
This skill should be used when modeling agent mental states with BDI concepts: beliefs, desires, intentions, RDF-to-belief transformations, rational agency traces, cognitive agents, BDI ontologies, and neuro-symbolic AI integration.
基于 SOC 职业分类
正在显示 SKILL.md
| name | code-execution |
| description | Execute Python code locally with marketplace API access for 90%+ token savings on bulk operations. Activates when user requests bulk operations (10+ files), complex multi-step workflows, iterative processing, or mentions efficiency/performance. |
Execute Python locally with API access. 90-99% token savings for bulk operations.
Use direct Python imports in Claude Code:
from execution_runtime import fs, code, transform, git
# Code analysis (metadata only!)
functions = code.find_functions('app.py', pattern='handle_.*')
# File operations
code_block = fs.copy_lines('source.py', 10, 20)
fs.paste_code('target.py', 50, code_block)
# Bulk transformations
result = transform.rename_identifier('.', 'oldName', 'newName', '**/*.py')
# Git operations
git.git_add(['.'])
git.git_commit('feat: refactor code')
If not installed: Run ~/.claude/plugins/marketplaces/mhattingpete-claude-skills/execution-runtime/setup.sh
fs): copy_lines, paste_code, search_replace, batch_copycode): find_functions, find_classes, analyze_dependencies - returns METADATA only!transform): rename_identifier, remove_debug_statements, batch_refactorgit): git_status, git_add, git_commit, git_pushBulk refactor (50 files):
from execution_runtime import transform
result = transform.rename_identifier('.', 'oldName', 'newName', '**/*.py')
# Returns: {'files_modified': 50, 'total_replacements': 247}
Extract functions:
from execution_runtime import code, fs
functions = code.find_functions('app.py', pattern='.*_util$') # Metadata only!
for func in functions:
code_block = fs.copy_lines('app.py', func['start_line'], func['end_line'])
fs.paste_code('utils.py', -1, code_block)
result = {'functions_moved': len(functions)}
Code audit (100 files):
from execution_runtime import code
from pathlib import Path
files = list(Path('.').glob('**/*.py'))
issues = []
for file in files:
deps = code.analyze_dependencies(str(file)) # Metadata only!
if deps.get('complexity', 0) > 15:
issues.append({'file': str(file), 'complexity': deps['complexity']})
result = {'files_audited': len(files), 'high_complexity': len(issues)}
✅ Return summaries, not data ✅ Use code_analysis (returns metadata, not source) ✅ Batch operations ✅ Handle errors, return error count
❌ Don't return all code to context ❌ Don't read full source when you need metadata ❌ Don't process files one by one
| Files | Traditional | Execution | Savings |
|---|---|---|---|
| 10 | 5K tokens | 500 | 90% |
| 50 | 25K tokens | 600 | 97.6% |
| 100 | 150K tokens | 1K | 99.3% |