Borrowing it
Nothing to install: this file belongs to fishzjp/qa-skills. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/fishzjp/qa-skills/main/AGENTS.mdgit clone --depth 1 https://github.com/fishzjp/qa-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/fishzjp/qa-skills/agents-md)<a href="https://agentmods.dev/instructions/fishzjp/qa-skills/agents-md"><img src="https://agentmods.dev/badge/instructions/fishzjp/qa-skills/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/fishzjp/qa-skills/agents-md"><img src="https://agentmods.dev/badge/instructions/fishzjp/qa-skills/agents-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.01261 | $0.01261 |
| Opus 5.5 | $0.00504 | $0.00504 |
| Sonnet 5.5 | $0.00252 | $0.00252 |
| Haiku 4.5 | $0.00126 | $0.00126 |
Grade A, and why
qa-skills AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
AGENTS.md — qa-skills 项目指令
本文件是维护者的迭代 SOP。"评测三轨道"及各处
eval/引用描述的是本地维护的评测链路 (2026-08-22 起不随公开仓库分发,公开仓库只含 skills 产品内容);公开侧证据链为按里程碑 随 Release 附带的增益矩阵快照。
使命(所有工作的对齐点)
安装 qa-skills 之后,对任何 agent、任何模型,在软件测试领域带来质的提升。 一切决策——改 skill、建评测、扩任务、发版本——的优先级判断都以这一句话为锚。
总目标章程("测试领域最强"的判定记分牌、审查-对比常设回路、达成判据)见本地
docs/GOAL.md(规划文件,不入库);本季度执行方案 = 补强方案 v1.2(Phase 0–4),
为章程记分牌的季度执行切片。
战略路径:迭代飞轮
skill bench 是驱动 qa-skills 迭代的手段,不是目的。标准回路:
跑 bench → 失败模式归类、定位 skill 短板 → 修改 skill → 复验(评测-修复闭环)
bench 的产出是"改哪里",不是总分。聚合数字(如覆盖 +8.7pp)用于发布,不用于指导迭代; 指导迭代的是逐 GT 点的失败明细及其归类(澄清缺失 / 边界遗漏 / 状态遗漏 / 断言强度不足 → 映射到 skill 的具体环节)。
北极星指标
跨模型 × 跨宿主的 On/Off 增益下限(min,不是均值)。
- 弱模型段位:追求可执行性、真实执行通过率等类别性质变(0.2→0.98 型)
- 强模型段位:覆盖增益趋零后,价值在流程纪律与系统行为(复审拦截、回归编排、接口一致)
- 对单一模型(当前 glm-5.2)迭代出的提升若不能跨模型复现,视为过拟合而非进步
评测三轨道(eval/)
- 模型矩阵(常设回归):标准回路固定 2–3 个段位的便宜模型(flash 级 + 中档), 强模型按里程碑跑。skill 修改的合入门槛之一:最弱模型上的增益不回退。
- in-situ 轨道:真实宿主会话中量三个数——触发正确率、装载文件集合、产出质量。 这是"任何 agent"的关键证据(注入式评测只证明内容值得装,不证明装了会生效)。
- 注入式轨道(现有 harness):内容质量哨兵,原样保留。
迭代纪律
- 只有类别性 / 大效应信号才触发 skill 变更;+2pp 级差异一律视为采样噪声 (n=3 的采样方差与单次普通修改的效应同量级,见澄清任务跨轮方向翻转案例)。
- 任务池轮换(30–50 任务抽半跑,季度换血),防 skill 对着题库优化。
- 迭代涉及的任务做双人 GT 复核——GT 错标会让飞轮朝错误方向持续优化,是最危险的失败模式。
- 日常真实使用中的触发失败 / 产出不佳案例,每周沉淀 2–3 个进任务池(dogfooding 反哺)。
- 里程碑版 Release 配增益矩阵快照(跨模型 On/Off 对比)与迭代记录——这是 skill bench 的自然形态, 也是"安装这个 skill 会发生什么"的纵向证据。
门禁入口(维护者本地 ≠ CI)
CI 与贡献者入口(CONTRIBUTING 自检)只覆盖已跟踪面:unittest discover tests -p "test_*.py"
- 两个校验器 + 安装器冒烟。
tests/test_harness.py(harness 单测,随 eval/ 本地维护、gitignore) 只能用 pytest 跑:python3 -m pytest tests/ -q(收集面 = 上述已跟踪用例 + harness 全部用例)。 改 skill / 加 skill 后本地合入前必须跑 pytest 口径——用例数一律以实际收集数为准、不写进文档 (2026-09-07 审查教训:qa-memory 合入使 harness 断言红了一次,unittest 入口永远看不到它, 红灯静默 9 天;同批 AGENTS 写死的用例计数在加测试的同一个提交里就过期了)。
报告与对外发布的定位
评测报告(eval/reports/)是里程碑快照,不是目标本身。对外数字发布前跑零成本污染三件套 (cutoff 核对 / n-gram 扫描 / canary),见本地维护的 eval/harness/CONTAMINATION.md(eval/ 为本地链路,不随公开仓库分发)。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +13 lines · +315 tokens per session a1fa990f54e7
- 1mo ago Changed · +19 tokens per session b472e5a393e9
- 2mo ago First seen · 52 lines · 927 tokens per session scan A 89e9542d7e04
qa-skills AGENTS.md is an instructions file published in the GitHub repository fishzjp/qa-skills (55 stars, last pushed 9d ago), licensed MIT. It adds 1,261 tokens to every session, about $0.0050 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
qmax-mcp copilot-instructions.md
Copilot instructions for Quality-Max/qmax-mcp, covering qualitymax qa workflow and adjacent tools.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).