zccz14

Required entry skill when you are an AIM Developer working on an existing AIM Task and must read the task via AIM Server, validate it against the latest baseline, execute the worktree and PR lifecycle, report field facts, and complete the bound OpenCode session.

Strongly recommended when you need to clarify questions, converge directions and uncertainties, do design or orienting work, or answer open-ended questions before deciding how to proceed and producing a task spec, especially when key clarifications could change the route.

Coordinator decision entry for AIM Task Pool maintenance; form an approvable POST /tasks/batch operations plan from Manager output, latest baseline facts, current Tasks, and rejected Task feedback before applying atomic Task Pool writes.

当用户已经明确批准创建新的 AIM Task，且需要把稳定的用户意图整理成候选 Task Spec 并提交到 POST /tasks 时使用。

Use when judging how README claims compare to the latest origin/main baseline and the coordinator needs direction signals without task creation or execution decisions.

AIM Manager guidance for evaluating README goals against the latest baseline, defining iteration direction and Dimensions, and preparing Coordinator-consumable evaluation signals without creating tasks or executing work.

Use when a GitHub repository needs its merge settings, default-branch ruleset required checks, or PR auto-merge behavior verified or standardized with gh.

当需要在 AIM 仓库内实现新功能、修复缺陷、调整行为或补回归保护，并且必须在写实现前执行真实测试优先纪律时使用。

当生成或探索性工作可能失败、需要多次尝试、涉及众多约束、需要验证，或会产生不应污染主 Agent 上下文的中间发现时，主动使用本技能。 本技能运行一个整洁的主智能体循环：主 Agent 只负责协调，将上下文导出到文件，派发创建者/审查者 SubAgent，将重试过程留在主上下文之外， 并且只接收工件路径以及 PASS / RETRY / FAILED 审查说明。

用于每一项编程任务，以控制流程复杂度。对于任何创建或修改代码的任务，还应使用 clean-agent 协调模式，让创建者和审查者 SubAgent 将 clean-code 作为共同的代码规范。它强制执行复杂度门禁，将每个分支视为新增的逻辑路径，要求有明确的业务理由，尤其适用于实现和代码审查。

每当用户要求编写、重写、缩短、澄清、审查、润色、重构或改写文档及任何面向人类的说明性文本时，请使用本技能。对于任何创建或修改文档的任务，还要使用 clean-agent 协调模式，使创建者和审查者 SubAgent 将 clean-doc 作为共同的沟通规范。它把信息转化为针对特定受众和决策情境的目标导向型沟通，优先考虑读者的语言和风格，并且只保留与决策相关的信息。

强制执行 CZ 的 Git 和 GitHub 交付风格，包括 gitmoji 提交、受保护的 main、仓库本地 worktree、仅 Squash 的 Auto Merge、拉取请求监控、GitHub Actions 验证和安全清理。每当 Codex 设置或更新 GitHub 仓库、创建功能 worktree、准备或发起拉取请求、跟进拉取请求直至合并并完成 Actions，或执行相关 Git 工作流时，都应使用本技能。

当代码运行缓慢、超时、被终止、消耗大量 CPU、执行长时间批处理作业、进行参数扫描、训练模型、处理大型数据集、运行模拟，或目标规模运行时间未知时，使用本技能。它指导智能体逐步探测运行时间，在可行时进行在线 N-T 进度插桩，采用固定超时采样，执行 N-T 扩展估算，并在尝试目标规模前作出全规模运行的通过/不通过决策。如果进行了优化，它要求保留慢速版本的规范输出，证明语义逐字节等价，并在通过并行投入更多 CPU 前优先消除无效工作。本技能不封装性能分析工具，也不提供通用优化方案；它负责判断应全规模运行、停止，还是转入性能分析和原则性优化。

当任务涉及困难推理、原因不明、无法解释的观测、失败的假设、模糊诊断、复杂系统、涌现行为、调查规划，或当前解释空间不完整的任何情形时，请使用此技能。此技能应用递归残差推理：将尚未解释的内容保留为显式残差，把该残差扩展为下一个推理空间，并生成一条从观测到结论、行动或有界停止原因的可追踪路径。

当创建、修改、解释、诊断或验证模型、规则、评分系统、预测、参数选择、机制假设或研究结论时，请使用此技能。它要求同时具备先验依据与实证验证来评估可靠性，并区分可靠结论与黑箱相关性、空洞叙事、事后线索及不可靠主张。

Use this skill when a task involves hard reasoning, unclear causes, unexplained observations, failed hypotheses, ambiguous diagnoses, complex systems, emergent behavior, investigation planning, or any situation where the current explanation space is incomplete. This skill applies Recursive Residual Reasoning: preserve what cannot yet be explained as an explicit residual, expand that residual into the next reasoning space, and produce a traceable path from observation to conclusion, action, or bounded stop reason.

Use proactively for generation or exploratory work when the task might fail, need multiple attempts, involve many constraints, require verification, or produce intermediate findings that should not pollute the main Agent context. This skill runs a clean main-agent loop: the main Agent coordinates only, exports context to files, dispatches creator/reviewer SubAgents, keeps retries out of the main context, and receives only artifact paths plus PASS / RETRY / FAILED audit notes.

Use this skill whenever code is slow, times out, gets killed, burns substantial CPU, runs long batch jobs, performs parameter scans, trains models, processes large datasets, runs simulations, or has unknown target-scale runtime. It guides progressive runtime probing, online N-T progress instrumentation when feasible, fixed-timeout sampling, N-T scaling estimation, and full-run go/no-go decisions before attempting target scale. If optimization happens, it requires preserving slow-version canonical output, proving byte-for-byte semantic equivalence, and prioritizing removal of wasted work before spending more CPU through parallelism. This skill does not wrap profiler tools and does not provide generic optimization recipes; it decides whether to run full scale, stop, or move into profiling and principled optimization.

Use this skill when creating, modifying, explaining, diagnosing, or validating a model, rule, scoring system, prediction, parameter choice, mechanism hypothesis, or research conclusion. It evaluates reliability by requiring both prior rationale and empirical validation, and by separating reliable conclusions from black-box correlations, empty stories, post-hoc leads, and unreliable claims.

Use this skill whenever the user asks to write, rewrite, shorten, clarify, review, polish, restructure, or adapt documentation or any human-facing explanatory text. It turns information into purpose-driven communication for a specific audience and decision context. It prioritizes the reader's native language and habitual language style, preserves only decision-relevant information, and avoids bloated background, excessive completeness, and domain-specific assumptions unless the current context explicitly defines them.

Use for every coding task to control flow complexity. Enforces a complexity gate that treats each branch as an added logical path requiring explicit business justification, especially during implementation and code review.

Use this skill whenever the user wants an interactive trading replay, order annotation viewer, strategy visualization, trade chart, buy/sell point review, or equity-curve-linked chart UI. It guides building a lightweight-charts based viewer with symbol/date navigation, keyboard shortcuts, URL parameters, candlestick main/sub panes, clickable equity curve navigation, and human-readable order rationale. Do not bind the implementation to any particular CSV schema, database, broker export, or project-specific trade format.