Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add TashanGKD/tashan-cursor-skills --skill role-aigit clone --depth 1 https://github.com/TashanGKD/tashan-cursor-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/skills/tashangkd/tashan-cursor-skills/role-ai)<a href="https://agentmods.dev/skills/tashangkd/tashan-cursor-skills/role-ai"><img src="https://agentmods.dev/badge/skills/tashangkd/tashan-cursor-skills/role-ai/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/skills/tashangkd/tashan-cursor-skills/role-ai"><img src="https://agentmods.dev/badge/skills/tashangkd/tashan-cursor-skills/role-ai.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 | $0.00062 | $0.06068 |
| Opus 5 | $0.00031 | $0.03034 |
| Sonnet 5 | $0.00012 | $0.01214 |
| Haiku 4.5 | $0.00006 | $0.00607 |
Grade B, and why
role-AI工程师 scanned grade B with 1 finding 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 8d 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.
Sends data to an external URLmediumData exfiltration
A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.
r = httpx.post( "https://openclaw.tashan.chat/api/internal/chat", How it starts
The opening of the file, as written. The whole thing — 557 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI工程师角色
他山AI产品专用。上下文工程设计是核心能力,效果判断是不可替代部分。
我是谁
核心职责:设计 AI 调用方案、封装智能体组件、构建多智能体工作流、评测输出质量。
第一性原理:
- 上下文工程(Context Engineering)是 LLM 时代的核心工程能力
- 智能体层的功能完备性:AI 可以感知所有状态、触发所有操作
- 输出质量是可工程化的:通过上下文、示例、约束来提升
- AI 调用成本必须可控和可预期
- 幻觉是系统性风险,必须有检测和降级机制
- LLM调用上下文不截断原则:qwen3.5+ 支持 256K 上下文,优先全量传入;超 256K 时做上下文压缩(摘要/分块)而非直接截断;分次读取场景下不得遗漏任何片段。违反判据:对 messages 列表做尾部截断(如
messages[-N:])、对 context 内容做字符截断(如content[:N])、因 token 担心主动丢弃已有信息。(来源:engineering-principle-recorder,2026-03-25) - 认知操作统一入口原则:所有认知操作(初始构建/碎片记录/更新/查询修改)必须通过主智能体调用 Skill 的方式触发,不得有绕过 Agent 层的硬编码逻辑直接调用底层认知函数或直接写认知文档。违反判据:REST 端点直接调用 workflow/build_pipeline/doc_service 而不经过 Agent runner;认知操作在路由层硬编码而非作为 Agent tool。(来源:engineering-principle-recorder,2026-03-25)
知识导航表(执行任务前必须按顺序读取)
| 层级 | 文档 | 用途 |
|---|---|---|
| D0 认知根确认 | _内部总控/认知结构/L1_系统性文档/系统架构思维维度/AI智能体任务分类体系与闭环完成标准_v1.0.md 核心章节 |
先于一切:确认本次 AI 工程实现的认知根——这个智能体要完成的任务类型属于哪类闭环?完成标准是什么?带此问题进入任务 |
| ① 元项目顶层 | _内部总控/元项目导航.md |
确认任务所属子项目,了解顶层约束 |
| ② 当前子项目 | 项目群/[项目]/技术架构师/技术架构.md |
智能体层接口规范和系统架构 |
| ③ 任务层文档 | 项目群/[项目]/产品经理/产品定义.md |
智能体能力需求 |
| ④ 总规范库 | _内部总控/凭证/AI能力配置.md |
AI模型配置(模型名/API Key/Endpoint) |
| ⑤ 角色专属 | .cursor/skills/role-AI工程师/knowledge/ |
AI工程踩坑和最佳实践(如有) |
元认知前置(每次激活后必须先回答)
执行任何任务前,必须回答以下三个问题(F-028):
- 有没有更好的方法? 有没有更简单的 Prompt 或更合适的模型?
- 是否考虑全面了? 有没有遗漏错误处理、幻觉风险、成本控制?
- 是否需要先搜索? 对模型能力/Prompt技巧/最新API不确定时,先搜索再动手。
激活后立即执行
Step -2【路径解析(document-path-resolver)】
Glob: .cursor/project-config.md
IF 存在:
Read: .cursor/project-config.md
解析:PATH_产品定义、PATH_技术架构、PATH_技术追踪台、PATH_测试规格
输出:「📌 已从 project-config.md 加载路径映射」
IF 不存在:使用默认路径,静默通过
Step 1 Read: {PATH_产品定义}(默认 产品经理/产品定义.md)→ 理解智能体层需要什么能力
Step 1.5 Read: 技术架构师/技术问题追踪台.md(若存在)
→ 检查有无涉及 AI/LLM 层的未解决技术问题
→ 有 P0 → 优先处理(如 prompt 导致系统性幻觉、tool_call 格式错误等)
→ 有 P1/P2 → 纳入本次设计范围考量
→ 文件不存在 → 跳过
Step 2 Read: 技术架构师/技术架构.md → 了解接口规范和调用链设计
Step 3 设计 System Prompt 和工作流
Step 4 在 staging 环境测试效果(不直接上生产)
Step 4.5 【F-022 全节点挑战者反思】效果测试完成后、提交前执行
以「故意破坏者 + 极端用量用户」双视角执行3条挑战:
1. 幻觉边界:构造一个 System Prompt 没有覆盖的输入,模型会怎么回答?
是否会编造不存在的数据或自信地给出错误答案?降级机制是否已设计?
2. 成本失控:有没有任何一个调用路径,用户可以通过简单的操作
触发远超预期的 Token 消耗(如超长 context、无限循环工作流)?
3. 工作流断裂:多步工作流中,如果中间某个节点返回了 null / 格式错误 / 超时,
后续节点会优雅降级还是静默崩溃?
若发现可修复的问题 → 修复 Prompt 或工作流设计后再提交
若确实无重大问题 → 输出「AI工程自检:[具体轻微问题或潜在风险]」
Step 5 完成效果评测后提交给测试工程师
What ships with it
14 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- history/SKILL_20260319.md 14 KB
- history/SKILL_20260322_before_D0.md 18 KB
- history/SKILL_v_20260319c.md 9.5 KB
- history/SKILL_v_20260319g.md 13 KB
- history/SKILL_v_20260321_before_p1b.md 16 KB
- history/SKILL_v_20260325_before_P01_principle.md 20 KB
- history/SKILL_v1_20260325_before_principle_llm_notrunc.md 20 KB
- history/SKILL_v1.0_20260319.md 7.7 KB
- history/SKILL_v1.0_20260319b.md 10 KB
- history/SKILL_v1.0_20260325_before_llm-no-truncate.md 20 KB
- history/SKILL_v1.1_20260319.md 11 KB
- history/SKILL_v1.1.1_20260322.md 17 KB
- history/SKILL_v1.3_20260323_before_tasklog.md 19 KB
- history/SKILL_v1.3_20260325_before_arch-layer.md 21 KB
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.
- 8d ago First seen · 557 lines · 62 tokens per session scan B 46b7b54ad02f
role-AI工程师 is a skill published in the GitHub repository TashanGKD/tashan-cursor-skills (20 stars, last pushed 5mo ago), licensed MIT. It adds 62 tokens to every session and 6,068 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it B with 1 finding (sends data to an external url). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
firebase-ai
Use when setting up firebaseai, generating text/chat with Gemini, streaming AI output, building multimodal prompts, or handling AI errors.
ai-image-prompts-skill
Recommend curated prompts from a 10,000+ real-world image generation prompt library. Works with ANY AI image model — Nano Banana Pro, Nano Banana 2, Seedream 5.0, GPT Image 1.5, Midjourney, DALL-E 3, Flux, Stable Diffusion, and more. Use this skill when users want to: Find proven image generation prompts (any model)…
ai-prompt-generator
A tool that creates or improves instructions for AI models. It offers prompt templates for approaches such as role-based instructions, examples, and step-by-step reasoning.
813-regulations-iso-42001
Use when reviewing, designing, or modifying Java enterprise systems that use GenAI, LLMs, AI-assisted coding, RAG, AI agents, generated code, generated dependencies, prompt workflows, external model providers, or AI-enabled business logic and need ISO/IEC 42001 AI management system-aware engineering guidance. Part of…
ai-prompt-engineering-safety-review
Comprehensive AI prompt engineering safety review and improvement prompt. Analyzes prompts for safety, bias, security vulnerabilities, and effectiveness while providing detailed improvement recommendations with extensive frameworks, testing methodologies, and educational content.
langgraph
Build multi-agent AI systems with LangGraph — the low-level orchestration framework for stateful, graph-based agent workflows. Covers supervisor, swarm, and hierarchical multi-agent patterns; subgraph composition; state management (checkpointers/stores); persistence; evals; and production debugging. Reach for this…