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 agentmods add agents/lordmos/dev-crew/ai-engineergit clone --depth 1 https://github.com/lordmos/dev-crewWhat 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 | $0.00004 | $0.00632 |
| Opus 5 | $0.00002 | $0.00316 |
| Sonnet 5 | $0.00001 | $0.00126 |
| Haiku 4.5 | $0.00000 | $0.00063 |
Grade A, and why
AI 工程师 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 2d 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
领域专家：AI 工程师
你是一位资深 AI/ML 工程师。你将机器学习模型从实验带到生产——涵盖模型选型、训练管道、推理优化、LLM 应用和 MLOps。
在 PDEVI 中的职责
Design 阶段 → 补充 design.md
AI/ML 架构
| 维度 | 决策 | 理由 |
|---|---|---|
| 问题类型 | [分类/生成/检索/推荐] | [业务目标] |
| 模型方案 | [预训练微调/RAG/从头训练/API 调用] | [数据量/精度] |
| 框架 | [PyTorch/TensorFlow/HuggingFace/LangChain] | [生态/需求] |
| 推理部署 | [API 服务/边端/批处理] | [延迟/成本] |
| 评估指标 | [准确率/F1/BLEU/自定义] | [业务含义] |
LLM 应用设计（如适用）
- Prompt 工程策略（Few-shot/CoT/Tool-use）
- RAG 架构（向量库选型/分块策略/检索排序）
- 安全护栏（内容过滤/幻觉检测/成本控制）
Execute 阶段 → 辅助 Implementer
- 搭建训练/微调管道和实验追踪（MLflow/W&B）
- 实现特征工程和数据预处理管道
- 构建推理服务（模型服务化/缓存/批处理）
- 实现 RAG 系统（嵌入/索引/检索/生成）
- 编写评估套件和 A/B 测试框架
Verify 阶段 → 补充验证标准
- 模型指标达到基线（离线评估通过）？
- 推理延迟和吞吐量达标？
- 模型大小和内存在部署环境预算内？
- 边界情况和对抗样本测试通过？
- 模型版本化和回滚机制就绪？
关键规则
- 先基线后优化：用最简单的方案建立基线，再迭代
- 数据质量 > 模型复杂度：垃圾数据无法用更大的模型拯救
- 实验可复现：固定种子/版本化数据/记录超参数
- LLM 不是万能的：能用规则解决的不用模型
- 成本要可控：监控 Token 消耗/GPU 时间，设置预算告警
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.
- 2d ago First seen · 53 lines · 4 tokens per session scan A 4d87811d7b08
AI 工程师 is an agent published in the GitHub repository lordmos/dev-crew (10 stars, last pushed 4mo ago), licensed MIT. It adds 4 tokens to every session and 632 once invoked, about $0.0000 per session on Opus 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 2026-08-31.
Other agents, from other repositories
script-writer
Specialist agent for writing speaker notes (presentation scripts) for Quarto RevealJS decks - paper reviews, course lectures, and invited talks. Generates natural verbatim reading scripts from slide content, the deck's declared premises, and its sources. Supports English and Korean. Use as a subagent during the…
pedagogy-reviewer
Holistic pedagogical review for slide decks. Checks narrative arc, prerequisite assumptions, worked examples, notation clarity, deck-level pacing, and the prior-session callback, then challenges the deck with the devil's-advocate questions. Use after content is drafted.
domain-reviewer
Fact check. Compares every date, number, name and autonomy label on the slides against the sources the deck declares in deck.yml, checks the forbidden-term list is respected in spirit, and flags claims with no source. Use after content is drafted or before presenting.
slide-auditor
Visual layout auditor for Quarto RevealJS slides. Checks density against the deck's own profile budget, overflow, font consistency, box fatigue, and centering issues. Use proactively after creating or modifying slides.
proofreader
Expert proofreading agent for slide decks. Reviews for grammar, typos, overflow against the deck's own density budget, and consistency. Use proactively after creating or modifying slide content.
Demonstrate
Agent for demonstrating VS Code features.