Lean Coding — Codex Python Beta

    1

    - Run the project's required checks and focused contract tests, then finish.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Lean Coding — Codex Python Beta

    Example session with this skill installed

    Recorded maintenance request (original Chinese; the machine-specific interpreter line is omitted):

    当前补全修订已解决多值选项与类似选项名的合法值归属,但新增加的 completion 元数据使原本可用的 Command.make_parser 覆写产生 AttributeError。请恢复既有解析工厂与 parse_args 标准三元组接口的调用能力;未提供新增观测字段不能使普通调用失败。保留现有多值补全、别名、重复选项、正常解析错误、用户补全异常及输入列表不变的行为,不回退原修复。

    • Read your context and instructions
    • Compiled the lean coding

    Recorded known-development example

    This is the complete final response from a real native Codex v7.1 development run, in its original Chinese. This is a previously known task, not a new production case or a performance comparison.

    Complete agent final response

    已完成维护需求。

    • src/click/core.py:为缺省的补全元数据使用默认值,恢复既有 make_parser 覆写和标准三元组接口兼容性。
    • 新增 test_parser_compatibility.py:覆盖覆写、多值补全、别名、重复选项、类似选项名的合法值、异常传播及输入列表不变。

    使用任务指定解释器运行 -B -m unittest discover -v,11 项测试全部通过;git diff --check 通过。内存还原旧逻辑后,新增测试成功复现 5 个兼容性错误。

    未完成项:无。下一步:审阅当前补丁。

    Production patch from the same run

    diff --git a/src/click/core.py b/src/click/core.py
    index b279e62..451c111 100644
    --- a/src/click/core.py
    +++ b/src/click/core.py
    @@ -1194,8 +1194,9 @@ class Command:
     
             parser = self.make_parser(ctx)
             opts, args, param_order = parser.parse_args(args=args)
    -        ctx._shell_complete_option = parser._shell_complete_option
    -        ctx._shell_complete_options = parser._shell_complete_options
    +        # Parser overrides may only implement the original three-value interface.
    +        ctx._shell_complete_option = getattr(parser, "_shell_complete_option", None)
    +        ctx._shell_complete_options = getattr(parser, "_shell_complete_options", True)
     
             for param in iter_params_for_processing(param_order, self.get_params(ctx)):
                 value, args = param.handle_parse_result(ctx, opts, args)
    

    Separate evaluator result

    The evaluator subsequently passed 28 focused contracts in ordinary Python and 28 in -O mode. These external checks are distinct from the 11 agent-authored local checks mentioned in the response. The frozen run also completed normally, and a reference-aware patch review found no confirmed defect. The package's other five known development tasks passed as well.

    This demonstrates a compatible fix on known material. It does not show that the skill caused the result, saves tokens, or generalizes to untouched projects. The full development report and patch are retained in the source repository:
    https://github.com/junsier9/lean-coding-agents/tree/main/evals/development-2026-10-28-v7-1

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Lean Coding — experimental beta

    Four concise decision rules for Codex maintaining existing Python code. Use it when a requested edit involves redundant validation, defensive copies, exception handling, fallback values, retries or wrappers.

    What it guides

    • Trace actual producers, callers, consumers and existing overrides before removing checks.
    • Preserve public behavior, accepted input protocols, errors, cancellation, transaction atomicity and cleanup.
    • Require a current requirement or demonstrated failure for each extra layer.
    • Run the project's required checks and focused contract tests, then finish.

    The instructions aim to reduce unnecessary machinery. A reduction in code, rework, execution time or token use has not been established.

    Evidence and limitations

    Package version 0.7.1-beta.1, using the unchanged v7.1 candidate. Six known development tasks completed successfully, with ordinary Python and -O checks. Two selected patches received reference-aware development review with no confirmed findings. These tasks informed the revision, ran once each and had no without-skill baseline. They are not unseen generalization or causal evidence.

    The original v7 development window failed artifact review with a compatibility regression, and the earlier v6 generalization window failed its zero-confirmed-defect gate. Those findings and scores remain preserved. The latest known regression fix does not establish production qualification.

    This is an early free experimental preview. The original beta quality gate remains unpassed. No stable speed or token-saving claim, industrial production qualification, or Claude validation is offered.

    Use

    In Codex, invoke $lean-coding with the requested Python edit and the project's actual acceptance criteria. Read-only explanations, read-only reviews, documentation-only changes and non-Python work are outside the current scope. Follow the project's existing tools and permissions.

    The package contains only SKILL.md, Codex UI metadata and the MIT license. It needs project file read/write and the existing Python/test shell. It adds no network service, bundled executable, extra dependency or paid API requirement. Model access follows your own Codex setup.

    Source, updates and feedback

    MIT source, complete evaluation limits, installation instructions and preserved failures: https://github.com/junsier9/lean-coding-agents

    Report reproducible issues at https://github.com/junsier9/lean-coding-agents/issues with the task, minimal code, expected/actual behavior and skill/model versions. Remove secrets and private project data. Future versions will include change notes; no update schedule is promised.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 20 days ago

    • Free with an Agensi account

    Software
    CodexPython

    Codex needs access to your Python project. It uses the project's existing interpreter and test tools.

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Reads and edits the requested project files and runs its existing checks. No additional network service, environment-variable access or account integration is required by this skill.

    Listed20 days ago

    What's inside

    Frequently Asked Questions

    Experimental Codex skill for lean Python maintenance: question redundant defenses while preserving real error contracts. Known regression checks passed; new-task quality, production readiness and speed/token benefits remain unproven.

    This skill uses the open SKILL.md standard. Use it through an MCP-compatible agent connected to Agensi, and check the compatibility information on this page for confirmed agents and required capabilities.

    Connect your AI agent to Agensi through MCP, then ask it to use Lean Coding — Codex Python Beta on your task. Your agent loads the latest approved version each time. Paid skills require a purchase or active subscription. Skill files are not available for download; your licence does not allow you to save, copy or export any part of a skill.

    New in Code Quality & Review

    Free