By summeronetwo
Claude Code plugin for competitive programming problem-setting workflows.
Readonly auditor for problem idea feasibility and verification readiness.
Readonly auditor before packaging, checks statement, tests, wrong solutions, and manifest consistency.
Readonly auditor for std/brute complexity and stress strategy.
Coordinates AutoCode problem creation and enforces the full validator-generator-checker workflow. Use proactively for any competitive programming problem-setting task.
Define and enforce project-wide quality standards for agent and skill documents. Use when creating, reviewing, or refactoring files under agents/ and skills/ to keep structure, terminology, and output contracts consistent.
Use when creating competitive programming problems with AutoCode MCP tools. Enforces the plugin workflow for problem statements, std/brute solutions, validators, generators, stress tests, final data verification, and Polygon packaging.
Use before coding to decide whether a problem idea is judgeable, implementable, and verifiable.
Validate statement samples and sample files for competitive programming problems. Ensures the expected outputs in problem statements match the actual solution output.
Audit std/brute assumptions with MCP evidence, including worst/average complexity risk and stress readiness.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
AutoCode 是面向竞赛编程出题人的 Claude Code plugin。
它不是单纯让 AI “写一道题”,而是把 AI 生成的题面、解法、校验器、生成器、对拍、测试数据和打包流程放进一条可验证、可审计、会阻止跳步的出题流水线。
仓库内部包含 autocode-mcp 这个 MCP server 实现,但默认分发和使用形态是 Claude Code plugin:安装后会同时获得工作流 Agent、Skills、Hooks 和 20 个 MCP 原子工具。
AI 可以快速给出题目想法和代码,但竞赛题真正难的是“可信”。常见翻车点通常不在第一眼能看出来的地方:
O(n^2) 被误当成能过大数据。AutoCode 的目标是把这些风险前置暴露,而不是等到出题完成后人工返工。
AutoCode 把“AI 出题”拆成一组必须通过的质量门禁:
| 风险 | AutoCode 的处理方式 |
|---|---|
| 题意不可判定、约束缺失 | autocode-idea-auditor 和 idea-feasibility 在实现前做只读审计 |
| 题面样例错误 | problem_validate 用解法验证题面样例与样例文件 |
| 标准解复杂度误判 | solution_analyze 输出复杂度、内存估计、风险提示和对拍建议 |
| 标准解或暴力解不可信 | solution_audit_std / solution_audit_brute 审计解法假设与 brute 能力 |
| std 有隐藏 bug | stress_test_run 用 sol/brute 多 profile 对拍 |
| 测试数据覆盖弱 | problem_generate_tests 生成 tiny/random/extreme/tle 多策略测试,并优先保证极限类比例 |
| type=3/type=4 语义重复 | problem_verify_tests 的 limit_semantics 检查极限/TLE 数据差异 |
| 错解未被杀 | problem_verify_tests 的 wrong_solution_kill 验证错解杀伤 |
| 工作流跳步 | Claude hooks 调用 scripts/workflow_guard.py,缺少前置步骤会直接拒绝工具调用 |
| 打包前状态不清 | autocode.json 记录题目契约,autocode-verify 可快速检查基础完整性 |
核心原则:AI 负责生成候选内容,AutoCode 负责让每一步必须被验证。
AutoCode 适合:
g++,推荐 GCC 10+testlib.h 已内置在 src/autocode_mcp/templates/,无需额外下载。
推荐通过 Claude Code marketplace 安装:
claude plugin marketplace add https://github.com/SummerOneTwo/autocode-marketplace.git
claude plugin install autocode@autocode-marketplace
安装后插件会启用:
autocode-workflowautocode-idea-auditor、autocode-solution-auditor、autocode-package-auditorautocode-workflow、idea-feasibility、solution-complexity-audit、stress-strategy、statement-audit、testdata-qualitySessionStart、PreToolUse、PostToolUseautocode-mcp安装完成后,可以直接描述你的出题目标:
用 AutoCode 创建一道竞赛编程题:给定数组,要求支持若干次区间查询。请先审计题意可行性,再按完整工作流生成题包。
工作流 Agent 会按门禁推进。如果你直接要求后置步骤,例如“现在打包”,但前置测试还没验证通过,hook 会拒绝并提示缺少哪一步。
非交互题主路径:
problem_create
-> solution_build(sol)
-> solution_build(brute)
-> solution_analyze / solution_audit_std / solution_audit_brute
-> validator_build(accuracy >= 0.9)
-> generator_build
-> stress_test_run(completed_rounds == total_rounds)
-> checker_build(需要特殊判题时)
-> problem_validate
-> problem_generate_tests
-> problem_verify_tests(passed)
-> problem_pack_polygon
交互题差异:
validator_build / checker_build
替换为 interactor_build
关键门禁:
brute 必须在 sol 之后构建。validator_build,且 accuracy >= 0.9。interactor_build。stress_test_run 必须完整跑完所有轮次。problem_generate_tests 前必须通过 problem_validate。problem_pack_polygon 前必须通过 problem_verify_tests,并满足门禁要求的结构化质量信号(如 limit_semantics、wrong_solution_kill、validator_check)。tests_verified 状态,必须重新验证。problem_create 会初始化标准题目目录:
<problem_dir>/
├── autocode.json
├── solutions/
│ ├── sol.cpp
│ └── brute.cpp
├── files/
│ ├── gen.cpp
│ ├── val.cpp
│ ├── checker.cpp
│ ├── interactor.cpp
│ └── testlib.h
├── statements/
│ ├── README.md
│ └── tutorial.md
└── tests/
├── 01.in
├── 01.ans / 01.out
└── .autocode_tests_manifest.json
autocode.json 是题目的可读契约,记录题名、是否交互、时空限制、题面路径、题解路径、解法角色和测试计划。示例:
npx claudepluginhub summeronetwo/autocodeInteractive LeetCode practice with learning-guided hints, solution submission, and AI-driven authentication via MCP server
In-context coding tutor for Claude Code. Learn from your real project with explanations, quizzes, diagnostics, and belt-based progression — locally and privately.
Programming as Theory Building guidelines for coding agents, grounded in Peter Naur's paper and focused on preserving program theory during code work.
Opt-in productivity coaching for Claude Code. Use for explicit PUA/try-harder requests, user frustration after repeated failures, requests to retry or change approach, passive/low-quality work complaints, completion checks, evidence requests, test/verification reminders, and Ding-style workplace process cues. Normal calm first-attempt requests are left alone.
Describe your goal, approve the spec, then step away — Claude and Codex loop together until it's right.
Auto-loop execution workflow with quality gates for Claude Code. Automatically decomposes tasks, implements code, runs tests, and iterates through quality gates until completion.