做什么
风险分级回波执行和代码证据协议. 跳过仅架构的工作,解释,保留合同的活字,和状态. 用于合同变更,调试,代码审查,实施计划,和Git突变;混合任务:只有这些部分.
技能库 智客分类:Agent 工作流 coding-protocol
风险分级回波执行和代码证据协议. 跳过仅架构的工作,解释,保留合同的活字,和状态. 用于合同变更,调试,代码审查,实施计划,和Git突变;混合任务:只有这些部分.
官方网址:skills.sh
先看中文介绍;官方 description 原文单独保留,不改写 SKILL.md。
风险分级回波执行和代码证据协议. 跳过仅架构的工作,解释,保留合同的活字,和状态. 用于合同变更,调试,代码审查,实施计划,和Git突变;混合任务:只有这些部分.
官方 description 未单独写出 Use when。按规范,代理会在用户任务与这段 description 的关键词匹配时激活本技能。
按 Agent Skills 渐进披露:启动时只加载 name 与 description(约 100 token);任务匹配后才读入整份 SKILL.md 正文;scripts/、references/、assets/ 仅在需要时再读。 本文件正文结构:Coding Protocol、Applicability Gate、Composition、1. Scale By Risk、2. Work From Evidence、3. Handle Material Uncertainty。
文件分析:除 SKILL.md 外,正文引用了 references/verification.md、references/rule-rationale.md,属于带资源的技能包,这些文件按需再读。
Risk-scaled repo execution and code-evidence protocol. Skip architecture-only work, explanation, contract-preserving prose, and status. Use for contract changes, debugging, code review, implementation plans, and Git mutation; mixed tasks: only those parts.
Coding ProtocolApplicability GateComposition1. Scale By Risk2. Work From Evidence3. Handle Material Uncertainty4. Keep Changes Small5. Preserve User Work6. Respect Local Context7. Verify Proportionally8. Report Only What MattersReferences
来源分类:skills.sh agent-skill
namecoding-protocoldescriptionreferences/verification.mdreferences/rule-rationale.md以下路径提取自原文;文件是否齐全请以来源仓库中的完整目录为准。
具体调用语法与可用工具以目标 Agent 客户端为准。 查看调用机制说明 ↗
先选择目标 Agent 和安装范围,保留技能包的附属文件,安装后检查客户端能否发现该技能。
该技能引用了附属文件,请从来源获取完整目录;仅复制 SKILL.md 可能缺少依赖。
复制安装指令给支持 Agent Skills 的代理,确认其中的目标目录与客户端匹配。
把 Agent Skill「coding-protocol」安装到我的项目:SKILL.md 原文与官方 description 见 https://zicq.com/zh/skills/skl-649b88310fd8eb85-%E7%BC%96%E7%A0%81%E5%8D%8F%E8%AE%AE.html 请存为 .cursor/skills/coding-protocol/SKILL.md 或 .claude/skills/coding-protocol/SKILL.md,frontmatter 的 name 与 description 保持原样,不要改写。 该技能还带 scripts/、references/、assets/ 等文件,请从 https://github.com/lencx/skills 取完整目录,不要只建一个 SKILL.md。
需要 Node.js 与 npx。先查看仓库技能列表,确认实际名称。
npx skills add 'https://github.com/lencx/skills' --list
npx skills add 'https://github.com/lencx/skills' --skill 'coding-protocol'
CLI 会交互选择目标 Agent,默认安装到项目;用户级安装使用 -g。先通过查看命令核对仓库内容,再用 npx skills list 检查已安装技能。
A low-friction protocol for reliable repository execution and code-evidence work. When applicable, run it in the background; surface only decisions, blockers, verification, and material risk.
Before any protocol step or reference read, select exactly one state in this order: Mixed, Execution, Evidence, Exit. Repository execution means implementing or planning a code, configuration, test, Git-state, or declared-contract change; code evidence means a repository-scoped diagnosis or code review. A contract-preserving prose-only edit is neither.
Repository evidence supplies local facts; repository instructions and contracts supply required checks. A focused workflow owns its method, vocabulary, artifact, professional judgment, and completion criterion. This protocol supplies only cross-cutting execution and code-evidence constraints: authorization, action-scope containment, work preservation, environment assumptions, evidence integrity, proportional verification, and truthful reporting.
Combine sources by concern; activation never supersedes another source, changes its completion criterion, or expands authority. Follow the host's established precedence when requirements conflict. Surface a material same-level conflict instead of silently choosing or accumulating incompatible requirements.
Use the focused workflow's completion criterion when one applies. Otherwise, use the narrowest safe method supported by repository evidence and, for multi-step work, name an observable completion criterion. Work as [action] -> [check], revising the route when evidence changes. Expose a plan only when it clarifies risk, coordination, or sequencing.
Scale effort with ambiguity, blast radius, and reversibility:
High-risk areas include auth, permissions, secrets, security, payments, data loss, schemas and migrations, public or shared contracts, concurrency, production configuration, dependency supply chains, and destructive actions. Risk raises the evidence bar, not the change scope.
Inspect relevant code, tests, types, documentation, contracts, and runtime output before claiming how the system works. Separate observations from assumptions when the difference matters. Never invent paths, APIs, checks, dependency behavior, performance, conventions, or project intent. A check passed only if it was run; when relevant evidence is unavailable, keep conclusions conditional.
Treat a runtime-contract mismatch as evidence to investigate, not as authority to widen accepted values or infer new semantics. Change the boundary only through the applicable authority and precedence rules.
Authority comes from the user and host, not skill activation. A request to explain, review, diagnose, or report does not authorize implementation. A request to fix, build, or change authorizes only its in-scope workspace mutation.
Automatic matching or loading grants no authority. A user request to execute a named skill authorizes only side effects that the request and the skill's declared purpose jointly make explicit, subject to host permissions. Mentioning, asking about, or comparing a skill authorizes no side effects by itself. Never infer adjacent commit, push, deploy, publish, external communication, or machine-wide operations.
Resolve low-risk ambiguity from evidence using the narrowest reasonable interpretation. Ask only when ambiguity affects correctness, safety, external behavior, user intent, irreversible work, or a high-risk area.
If the requested outcome, stated facts, or mechanism conflicts with observed evidence, show the mismatch before acting. Neither comply blindly nor silently substitute a materially different result. When no user can answer, park the ambiguous item and continue only with the unambiguous remainder. Keep work under uncertainty local and reversible.
Evidence stays read-only: bound inspection and conclusions to the requested code-evidence judgment. Execution and the execution portion of Mixed make the smallest complete change consistent with the authorized request, repository contracts, and focused workflow. Match local patterns. Every changed line must trace to one of those sources: avoid unrequested features, abstractions, speculative paths, formatting churn, dependencies, and unrelated fixes. Necessary call-site, invariant, migration, recovery, and verification changes remain in scope.
An execution slice does not define or expand the intended target. Take target scope from the request or an authoritative source. If unresolved scope would materially change the implementation, its promises, or required evidence, ask one decision-changing question or keep the change reversible and broader reuse unclaimed. A declared broader target may require representative evidence; it does not authorize migrating the remainder.
For that execution work, prefer a simpler supported solution when the mechanism was only a suggestion; preserve it when it is material to user intent. Refactor only as needed for a safe completion. Remove artifacts made obsolete by this task; report unrelated issues and leave pre-existing dead code alone.
Preserve user changes outside the task. Before broad edits, inspect the relevant diff; treat unfamiliar modifications as user-owned unless evidence ties them to this task. Do not overwrite, delete, move, reformat, or revert them.
Task necessity is not destructive authorization. Discarding work, resetting state, rewriting history, deleting broad or unresolved targets, and machine-wide changes require explicit authority. An in-scope edit may remove a precisely identified obsolete file when evidence establishes necessity and the recovery risk is understood. Once authority, target, and recovery are resolved, destructiveness alone is not a blocker.
Inspect relevant local signals before depending on package managers, tools, dependencies, network, credentials, ports, services, or writable paths. Prefer project-local commands and existing dependencies. Add dependencies, change tooling or lockfiles, start services, use network, or alter machine-wide state only when the task needs it and project evidence supports it.
Run checks required by the user, repository, or focused workflow; add the cheapest evidence proportional to risk. A green gate proves behavior only when it exercises the changed path.
Get to green honestly. A new or changed evidentiary check must be capable of failing, and expected behavior must come from the request, a repository contract, or an explicit characterization. Never turn a failure green through weaker assertions, skipped checks, unjustified suppression, or scope escape.
Diagnose before widening the change. If attempts thrash, stop with the diagnosis. Report every skipped, blocked, or failed verification and its residual risk.
Produce the artifact and handoff required by the task or focused workflow. Report what changed or was concluded, what was verified, and what remains unverified, blocked, or risky. Before declaring completion, re-check the primary criterion and every original requirement; name anything dropped, deferred, or reinterpreted. Keep the response concrete, omit routine internal process, and do not overstate confidence.
After the applicability gate, load only the reference that applies:
references/verification.md — only when designing or changing an evidentiary check, establishing a negative control, relying on representation-level evidence, or considering a suppression.references/rule-rationale.md — only before changing or auditing this protocol's rules; keep every failure-mode-to-section mapping accurate.Load neither reference on Exit.
Agent 工作流
创造有效技能指南。 当用户想创造出新的技能(或更新现有的技能),以专业知识,工作流程,或工具集成来扩展克洛德的能力时,应该使用这种技能.
Agent 工作流
使用ClawdHub CLI搜索,安装,更新并发布从taladhub.com的代理技能. 需要获取苍蝇上的新技能时使用,将安装的技能同步到最新版本或特定版本,或者发布 npm-instainddhub CLI 的新/更新的技能文件夹.
Agent 工作流
管弦乐团多代理团队,任务设定周期,交接协议,审查工作流程. 使用时间: (1)建立2+特派员队伍,具有不同专业,(2)确定任务路线和生命周期(收录框_ spec_建设_审查_完成),(3)在特派员之间制定交接协议,(4)建立审查和质量关口,(5)管理特派员之间的交流和文物共享.
Agent 工作流
Spec-first,TDD,子代理驱动的软件开发工作流程. 当:(1)构建任何新功能或应用——触发脑暴_计划_子代理执行回路,(2)调试出一个bug或测试失败——触发系统性的根起过程,(3)用户说"让我们构建","帮助我计划","我想添加X",或"这个被打破",(4)完成一个功…