Usage-based billing brief【免费下载链接】agent-skillsProduction-grade engineering skills for AI coding agents.项目地址: https://gitcode.com/GitHub_Trending/agentskill/agent-skillsThe product currently charges one flat monthly price. Leadership wants usage- based billing next quarter, but usage has not been defined. Candidate meters include API requests, processed records, and successful jobs.Known constraints:Existing customers need a migration path.Billing events must be auditable and idempotent.Late-arriving events occur for up to seven days.Finance requires invoice reconciliation.Unknowns include pricing tiers, free allowances, meter ownership, correction rules, customer-facing usage visibility, and regional tax behavior. Produce a spec and surface these decisions; do not implement them by assumption.它不是产品文档而是 [evals/cases/spec-driven-development.json](https://link.gitcode.com/i/182082249c4113aaa4bfa947e41fa669) 中行为评测的输入素材——评测 runner 会把它物化进一个一次性工作区交给被测 Agent 处理。它的短小恰恰是设计目的整个文档由三个要素构成 ### 1. 未定义的usage与三个候选计量口径 简报明确指出产品当前是固定月费管理层要求下季度转向按量计费但 usage has not been defined并给出三个候选计量口径meter**API 请求数、处理记录数、成功任务数**。这模拟了真实需求中最典型的危险状态方向明确要按量计费、核心定义缺失计量什么。一个遵循 spec-driven-development 技能的 Agent 应当把它写进规格的 Open Questions而不是自行挑选successful jobs就开始设计数据模型。 ### 2. 四条已知约束 - **存量客户需要迁移路径**Existing customers need a migration path——规格必须回答老客户如何从固定月费过渡 - **计费事件必须可审计且幂等**auditable and idempotent——这是事件流设计的硬性质同一事件重放不能重复计费 - **迟到事件最长可达 7 天**Late-arriving events occur for up to seven days——这一条直接决定窗口/结算周期的设计规格必须说明 7 天内的迟到事件如何并入已生成的账期 - **财务要求可对账的发票**invoice reconciliation——事件计量与最终发票之间必须存在可追溯链路。 这四条约束是可验证的规格要点后续无论规格怎么写这四条都必须被覆盖评测的 expectations[] 因此可以客观判断 Agent 是否遗漏。 ### 3. 六个未知决策项与一句明确的禁令 Unknowns 列出了 **定价层级、免费额度、计量口径归属、纠错规则、客户侧用量可见性、区域税务行为** 六项并以最后一句收尾 Produce a spec and surface these decisions; do not implement them by assumption. 产出一份规格并把这些决策显式暴露出来不要靠假设去实现它们。 这句是整份夹具的灵魂它与 [skills/spec-driven-development/SKILL.md](https://link.gitcode.com/i/70889c7902d7b7d0466b54e1c39c75e3) 中Surface assumptions immediately立即暴露假设Dont silently fill in ambiguous requirements不要静默补全模糊需求的指令直接对应——评测要检验的就是 Agent 能否把未知转化为向人提问而不是编一个合理默认值继续干。 ## 二、夹具如何接入 Tier-3 行为评测 ### 评测案例prompt、fixture 与期望断言 [billing-brief.md](https://link.gitcode.com/i/01d885e66717c1bc28535c470eaf8ce0) 由 [evals/cases/spec-driven-development.json](https://link.gitcode.com/i/182082249c4113aaa4bfa947e41fa669) 中 id 为 1 的评测消费该文件第 41–55 行 json { id: 1, prompt: We want usage-based billing. Produce the spec before any implementation., expected_output: A SPEC.md covering objectives, behavior, data model, testing strategy, and explicit boundaries, files: [spec-driven-development], expectations: [ The spec includes explicit boundaries and non-goals, Ambiguities are surfaced as questions rather than silently resolved, A testing strategy is part of the spec, No implementation code is written ] }几个字段的对应关系值得注意files: [spec-driven-development]指向evals/fixtures/spec-driven-development/目录即billing-brief.md所在目录。执行型评测kind 默认execution要求非空files[]路径相对于evals/fixtures/四条expectations[]是可判定陈述verifiable statements逐条对应夹具的设计第一条对应约束清单要落成 boundaries/non-goals第二条对应 Unknowns 与那句不要靠假设实现的禁令第三条要求测试策略写进规格呼应简报中可审计、幂等等性质可被测试验证第四条No implementation code is written直接检验do not implement的指令是否被遵守。runner 的执行链路评测的运行机制见 evals/README.md 与 scripts/run-evals.js。--behavioral模式下物化夹具files[]指向的真实项目输入从evals/fixtures/拷贝进一次性 git 仓库并作为基线提交fixture 路径解析见resolveFixturePathscripts/run-evals.js会拒绝绝对路径与逃逸出工作区的路径执行headlessclaude以--permission-mode acceptEdits加预批准工具列表Read,Glob,Grep,Edit,Write,Bash,WebFetch,WebSearch见 scripts/run-evals.js运行执行器超时常量为 15 分钟scripts/run-evals.js评分完整--output-format stream-json --verbose执行轨迹含工具调用被作为不可信数据加围栏后通过 stdin 喂给评分器轨迹可能达数 MBargv 会撞 OS 参数长度上限评分器输出必须校验为 JSON 后才写入evals/results/已 gitignore。对billing-brief.md这个案例评分器会检查规格是否把usage 未定义列成待决问题、七条类约束迁移、幂等、迟到 7 天、对账是否进入规格、Agent 是否真的没写实现代码。# Tier 3 — behavioral, runs each eval through headless claude, then grades it node scripts/run-evals.js --behavioral spec-driven-development # 消耗 tokens node scripts/run-evals.js --behavioral spec-driven-development --dry-run # 只打印计划三、与 Tier-2 路由评测的衔接billing 相关 prompt 的触发验证同一份案例文件还定义了触发trigger评测其中与 billing 主题强相关的正例 prompt 包括{ prompt: Write a spec for the new billing feature before we write any code, top_k: 3 }负例则声明了owner把不该被本技能接走变成成对路由测试例如Break the approved spec into ordered implementation tasks 的 owner 是planning-and-task-breakdownowner 必须在排序中压过本技能否则算错误——见 scripts/run-evals.js 的实现。Tier-2 用词干化 TF-IDF 对全部 25 个技能的description做词法近似评分描述两两余弦相似度 ≥75% 报错、≥50% 警告scripts/run-evals.js。在当前仓库上实际运行node scripts/run-evals.js的结果为Running skill evals across 25 skills, 25 case files 140 checks passed — 0 error(s), 0 warning(s) trigger rank-1 rate: 86% (76/88 positive prompts rank their skill first) PASSED也就是说billing spec这类真实用户说法确实能把 spec-driven-development 排进 top-k且整个描述目录没有近似碰撞。四、被评测的技能本体规格文档必须覆盖的六要素夹具检验的技能流程定义在 skills/spec-driven-development/SKILL.md。billing-brief.md的每个元素都映射到规格模板的某个部分简报元素规格落点usage 未定义 三个候选 meterOpen Questions——作为待人类决策的问题列出四条已知约束迁移/幂等/迟到 7 天/对账Success Criteria与Boundaries——可测试的完成条件六项 UnknownsOpen Questions——Produce a spec and surface these decisions 的直接执行do not implement by assumptionPhase 4 之前的门禁——规格未获批不进入实现该技能的核心是四阶段门禁工作流SPECIFY ──→ PLAN ──→ TASKS ──→ IMPLEMENT │ │ │ │ ▼ ▼ ▼ ▼ Human Human Human Human reviews reviews reviews reviews【免费下载链接】agent-skillsProduction-grade engineering skills for AI coding agents.项目地址: https://gitcode.com/GitHub_Trending/agentskill/agent-skills创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考 Usage-based billing brief
Usage-based billing brief【免费下载链接】agent-skillsProduction-grade engineering skills for AI coding agents.项目地址: https://gitcode.com/GitHub_Trending/agentskill/agent-skillsThe product currently charges one flat monthly price. Leadership wants usage- based billing next quarter, but usage has not been defined. Candidate meters include API requests, processed records, and successful jobs.Known constraints:Existing customers need a migration path.Billing events must be auditable and idempotent.Late-arriving events occur for up to seven days.Finance requires invoice reconciliation.Unknowns include pricing tiers, free allowances, meter ownership, correction rules, customer-facing usage visibility, and regional tax behavior. Produce a spec and surface these decisions; do not implement them by assumption.它不是产品文档而是 [evals/cases/spec-driven-development.json](https://link.gitcode.com/i/182082249c4113aaa4bfa947e41fa669) 中行为评测的输入素材——评测 runner 会把它物化进一个一次性工作区交给被测 Agent 处理。它的短小恰恰是设计目的整个文档由三个要素构成 ### 1. 未定义的usage与三个候选计量口径 简报明确指出产品当前是固定月费管理层要求下季度转向按量计费但 usage has not been defined并给出三个候选计量口径meter**API 请求数、处理记录数、成功任务数**。这模拟了真实需求中最典型的危险状态方向明确要按量计费、核心定义缺失计量什么。一个遵循 spec-driven-development 技能的 Agent 应当把它写进规格的 Open Questions而不是自行挑选successful jobs就开始设计数据模型。 ### 2. 四条已知约束 - **存量客户需要迁移路径**Existing customers need a migration path——规格必须回答老客户如何从固定月费过渡 - **计费事件必须可审计且幂等**auditable and idempotent——这是事件流设计的硬性质同一事件重放不能重复计费 - **迟到事件最长可达 7 天**Late-arriving events occur for up to seven days——这一条直接决定窗口/结算周期的设计规格必须说明 7 天内的迟到事件如何并入已生成的账期 - **财务要求可对账的发票**invoice reconciliation——事件计量与最终发票之间必须存在可追溯链路。 这四条约束是可验证的规格要点后续无论规格怎么写这四条都必须被覆盖评测的 expectations[] 因此可以客观判断 Agent 是否遗漏。 ### 3. 六个未知决策项与一句明确的禁令 Unknowns 列出了 **定价层级、免费额度、计量口径归属、纠错规则、客户侧用量可见性、区域税务行为** 六项并以最后一句收尾 Produce a spec and surface these decisions; do not implement them by assumption. 产出一份规格并把这些决策显式暴露出来不要靠假设去实现它们。 这句是整份夹具的灵魂它与 [skills/spec-driven-development/SKILL.md](https://link.gitcode.com/i/70889c7902d7b7d0466b54e1c39c75e3) 中Surface assumptions immediately立即暴露假设Dont silently fill in ambiguous requirements不要静默补全模糊需求的指令直接对应——评测要检验的就是 Agent 能否把未知转化为向人提问而不是编一个合理默认值继续干。 ## 二、夹具如何接入 Tier-3 行为评测 ### 评测案例prompt、fixture 与期望断言 [billing-brief.md](https://link.gitcode.com/i/01d885e66717c1bc28535c470eaf8ce0) 由 [evals/cases/spec-driven-development.json](https://link.gitcode.com/i/182082249c4113aaa4bfa947e41fa669) 中 id 为 1 的评测消费该文件第 41–55 行 json { id: 1, prompt: We want usage-based billing. Produce the spec before any implementation., expected_output: A SPEC.md covering objectives, behavior, data model, testing strategy, and explicit boundaries, files: [spec-driven-development], expectations: [ The spec includes explicit boundaries and non-goals, Ambiguities are surfaced as questions rather than silently resolved, A testing strategy is part of the spec, No implementation code is written ] }几个字段的对应关系值得注意files: [spec-driven-development]指向evals/fixtures/spec-driven-development/目录即billing-brief.md所在目录。执行型评测kind 默认execution要求非空files[]路径相对于evals/fixtures/四条expectations[]是可判定陈述verifiable statements逐条对应夹具的设计第一条对应约束清单要落成 boundaries/non-goals第二条对应 Unknowns 与那句不要靠假设实现的禁令第三条要求测试策略写进规格呼应简报中可审计、幂等等性质可被测试验证第四条No implementation code is written直接检验do not implement的指令是否被遵守。runner 的执行链路评测的运行机制见 evals/README.md 与 scripts/run-evals.js。--behavioral模式下物化夹具files[]指向的真实项目输入从evals/fixtures/拷贝进一次性 git 仓库并作为基线提交fixture 路径解析见resolveFixturePathscripts/run-evals.js会拒绝绝对路径与逃逸出工作区的路径执行headlessclaude以--permission-mode acceptEdits加预批准工具列表Read,Glob,Grep,Edit,Write,Bash,WebFetch,WebSearch见 scripts/run-evals.js运行执行器超时常量为 15 分钟scripts/run-evals.js评分完整--output-format stream-json --verbose执行轨迹含工具调用被作为不可信数据加围栏后通过 stdin 喂给评分器轨迹可能达数 MBargv 会撞 OS 参数长度上限评分器输出必须校验为 JSON 后才写入evals/results/已 gitignore。对billing-brief.md这个案例评分器会检查规格是否把usage 未定义列成待决问题、七条类约束迁移、幂等、迟到 7 天、对账是否进入规格、Agent 是否真的没写实现代码。# Tier 3 — behavioral, runs each eval through headless claude, then grades it node scripts/run-evals.js --behavioral spec-driven-development # 消耗 tokens node scripts/run-evals.js --behavioral spec-driven-development --dry-run # 只打印计划三、与 Tier-2 路由评测的衔接billing 相关 prompt 的触发验证同一份案例文件还定义了触发trigger评测其中与 billing 主题强相关的正例 prompt 包括{ prompt: Write a spec for the new billing feature before we write any code, top_k: 3 }负例则声明了owner把不该被本技能接走变成成对路由测试例如Break the approved spec into ordered implementation tasks 的 owner 是planning-and-task-breakdownowner 必须在排序中压过本技能否则算错误——见 scripts/run-evals.js 的实现。Tier-2 用词干化 TF-IDF 对全部 25 个技能的description做词法近似评分描述两两余弦相似度 ≥75% 报错、≥50% 警告scripts/run-evals.js。在当前仓库上实际运行node scripts/run-evals.js的结果为Running skill evals across 25 skills, 25 case files 140 checks passed — 0 error(s), 0 warning(s) trigger rank-1 rate: 86% (76/88 positive prompts rank their skill first) PASSED也就是说billing spec这类真实用户说法确实能把 spec-driven-development 排进 top-k且整个描述目录没有近似碰撞。四、被评测的技能本体规格文档必须覆盖的六要素夹具检验的技能流程定义在 skills/spec-driven-development/SKILL.md。billing-brief.md的每个元素都映射到规格模板的某个部分简报元素规格落点usage 未定义 三个候选 meterOpen Questions——作为待人类决策的问题列出四条已知约束迁移/幂等/迟到 7 天/对账Success Criteria与Boundaries——可测试的完成条件六项 UnknownsOpen Questions——Produce a spec and surface these decisions 的直接执行do not implement by assumptionPhase 4 之前的门禁——规格未获批不进入实现该技能的核心是四阶段门禁工作流SPECIFY ──→ PLAN ──→ TASKS ──→ IMPLEMENT │ │ │ │ ▼ ▼ ▼ ▼ Human Human Human Human reviews reviews reviews reviews【免费下载链接】agent-skillsProduction-grade engineering skills for AI coding agents.项目地址: https://gitcode.com/GitHub_Trending/agentskill/agent-skills创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考