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 mthines/agent-skills --skill ai-engineeringgit clone --depth 1 https://github.com/mthines/agent-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/mthines/agent-skills/ai-engineering)<a href="https://agentmods.dev/skills/mthines/agent-skills/ai-engineering"><img src="https://agentmods.dev/badge/skills/mthines/agent-skills/ai-engineering/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/mthines/agent-skills/ai-engineering"><img src="https://agentmods.dev/badge/skills/mthines/agent-skills/ai-engineering.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Prompt Injection · line 162 Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.Fix: Remove the large whitespace padding (blank-line blocks or long space runs) and review any content hidden below or to the right of it. Keep skill files compact and reviewable so no instructions can be
- medium Excessive Agency · line 262 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
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 | $0.00067 | $0.03141 |
| Opus 5.5 | $0.00027 | $0.01256 |
| Sonnet 5.5 | $0.00013 | $0.00628 |
| Haiku 4.5 | $0.00007 | $0.00314 |
Grade A, and why
ai-engineering 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 7d 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.
How it starts
The opening of the file, as written. The whole thing — 257 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Engineering
Prescriptive guidance for building and reviewing LLM/AI applications. Thirteen orthogonal concerns — load only the rules the current task needs.
This
SKILL.mdis a thin index. Detailed rules live inrules/*.mdand load on demand. Curated source URLs live inreferences/primary-sources.md. Date-flagged changes since 2024 live inreferences/recent-changes.md. Literal scaffolding lives intemplates/.
Mode Detection
Parse $ARGUMENTS (first token) and detect the mode:
| Mode | Default | Trigger |
|---|---|---|
guide |
yes | Default. Open question ("how should I …", "what's the best way to …"). |
review |
$0 == "review", or a file/path is supplied as $ARGUMENTS. |
|
design |
$0 == "design", or "scaffold a prompt / system prompt / eval". |
State the detected mode and the area(s) in scope before continuing:
Mode: review
Areas: prompt-writing, system-prompt-design
Targets: src/agents/triage.ts (system prompt at L24-78)
Area Routing
Map the user's request to one or more rule files. Load only the rules listed for the matched area(s).
| Area | Rule file | Load when |
|---|---|---|
| Writing user prompts | rules/prompt-writing.md |
"improve this prompt", few-shot questions, output format, CoT, structured outputs. |
| Designing system prompts | rules/system-prompt-design.md |
Persona, tool docs, ordering, refusals, agent stop conditions. |
| Token cost / latency | rules/token-optimization.md |
Prompt caching, model routing, batching, streaming, max_tokens. |
| Multimodal | rules/multimodal.md |
Image/audio/PDF inputs, vision-vs-OCR, voice agents, image token costs. |
| Retrieval-augmented generation | rules/rag.md |
Chunking, embeddings, hybrid search, reranking, query rewriting. |
| Agents & tool use | rules/agents-and-tools.md |
Tool schemas, agent loops, parallel tool calls, error recovery, workflow vs agent. |
| Resilience | rules/resilience.md |
Rate limits (429), retries with jitter, circuit breakers, fallback chains, timeouts, idempotency. |
| Memory & long-running state | rules/memory-and-state.md |
Conversation summarisation, structured memory, vector memory, memory tools, compaction. |
| Model migration & versioning | rules/model-migration.md |
Pin snapshots vs aliases, A/B a new model, deprecations, cross-provider migration, rollback. |
| Evaluation | rules/evals.md |
Golden sets, LLM-as-judge, regression CI, error analysis. |
| Testing (engineering) | rules/testing.md |
Unit/integration tests, mocks, VCR cassettes, snapshot tests, CI cost discipline. |
| Safety & guardrails | rules/safety-and-guardrails.md |
Prompt injection, jailbreaks, output validation, PII, scope control. |
| Observability & versioning | rules/observability-and-versioning.md |
Tracing, prompts-as-code, A/B releases, rollback. |
What ships with it
20 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.
- lens.md 2.8 KB
- references/primary-sources.md 7.3 KB
- references/recent-changes.md 3.7 KB
- rules/agents-and-tools.md 6.3 KB
- rules/evals.md 7.7 KB
- rules/memory-and-state.md 8.0 KB
- rules/model-migration.md 7.1 KB
- rules/multimodal.md 7.7 KB
- rules/observability-and-versioning.md 10 KB
- rules/prompt-writing.md 7.6 KB
- rules/rag.md 5.9 KB
- rules/resilience.md 7.4 KB
- rules/safety-and-guardrails.md 8.3 KB
- rules/system-prompt-design.md 7.3 KB
- rules/testing.md 12 KB
- rules/token-optimization.md 7.1 KB
- templates/eval-rubric.md 1.9 KB
- templates/golden-set.md 3.5 KB
- templates/system-prompt-skeleton.md 2.0 KB
- templates/tool-description.md 3.2 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.
- 7d ago Changed · -6 lines · -94 tokens per session 9d541fd17e0a
- 1mo ago Changed c64572108dee
- 1mo ago First seen · 263 lines · 161 tokens per session scan A b61e82eb2a6c
ai-engineering is a skill published in the GitHub repository mthines/agent-skills (20 stars, last pushed today), licensed MIT. It adds 67 tokens to every session and 3,141 once invoked, about $0.0003 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 2026-08-31.
Other skills, from other repositories
genkit-go-flows
Guides building AI features with Genkit for Go - flows, prompts, structured generation, tool calling, and middleware. Use when defining a Genkit flow, writing a prompt, calling genkit.Generate/GenerateData, defining a tool, debugging Genkit output with traces, or asked how Genkit differs from Google's Agent…
genai-hub-orchestration
Configure and use the SAP AI Core Orchestration service: build prompt templates, add content filtering, set up document grounding (RAG) pipelines, and call models via the Harmonized API. WHEN: user asks about the orchestration service, Harmonized API, prompt templating with variables, content filtering, grounding, RAG…
ai-llm-skills-guide
Guide for AI Agents and LLM development skills including RAG, multi-agent systems, prompt engineering, memory systems, and context engineering.
ai-product
Every product will be AI-powered. The question is whether you'll build it right or ship a demo that falls apart in production.
flockion_engineering_ai
Forces the simplest, shortest, safest AI solution that actually works. Lazy senior-engineer pragmatism for LLM features, agents, RAG, tool-calling, and prompts. Use a prompt before a chain, retrieval before fine-tuning, deterministic code before an agent, and one tool before a tool framework. Keep critical business…
ai-teacher
A guide to teaching about artificial intelligence, including how large language models work, prompt writing, AI agents, tools, and AI ethics.