One agent, three runtimes: porting a CSA agent to Microsoft Scout and Foundry Local
Most of my posts here are about Azure infrastructure lessons from customer engagements. This one is a little different — it's a real‑world engineering lesson from something I built to run my own practice. In my role as a Senior Cloud Solution Architect (CSA), I'm part of a grass-roots organic development team for an internal persona‑driven productivity agent called CSA‑Sherpa. It runs my daily rhythm: a morning briefing, a running logbook of wins and blockers, pipeline and timekeeping summaries, and reporting/exports. It started life in the GitHub Copilot CLI. But over the last few months two things changed the ground under it: Microsoft Scout arrived as a managed cloud agent with native tooling, scheduling, and memory; and Foundry Local made it realistic to run a capable model entirely on‑device on a Copilot+ PC's NPU — no cloud round‑trip at all. That raised a question I think a lot of people building agents will eventually ask: If I designed the framework well, can I change how the model runs without rewriting the agent? To find out, I stood the same agent up in three runtimes, then wrote a whitepaper and a comparison deck measuring what actually changed. This post explains: How one shared, deterministic core made three very different runtimes comparable What the three ports — Copilot CLI, Scout‑native, and Foundry Local (on‑device NPU) — actually took What the analysis showed, and a simple decision framework for which runtime to use when The part that stayed the same: a deterministic core The whole exercise only works because all three implementations load the same behavioral core: Agent definition — persona, behavioral rules, intent routing, workflow dispatch Instructions — conventions, session bootstrap, change‑management rules Skill library — one procedure file per workflow (morning briefing, logbook, pipeline, timekeeping, impact, ops, export…) A deterministic validation contract — schema, formatting, and privacy validators plus a post‑save enforcement chain That last point is the whole thesis: reliability belongs in code, not in the prompt. Rather than asking the model to "remember" to validate its output, a real gate (a validation step → a post‑save enforcement chain → index regeneration) enforces it every single run. This wasn't my idea in a vacuum — it follows the enterprise prompt‑engineering principles Kathiravan Thangavelu lays out in his article Prompt Engineering for Enterprise AI: Why Reliability Matters: keep deterministic logic in code, prefer schema‑driven / structured output over prompt‑enforced formatting, and replace "before you answer, verify that…" mental checklists with real machine validation. My validation gate is that principle in practice. And because that contract is identical across all three runtimes, I'm comparing three ways to execute one product — not three different products. The deterministic payoff: faster and cheaper Retrofitting those principles into the agent — moving work out of the model and into deterministic scripts — is the single change that paid off the most, on two axes at once: Faster. Letting code (not the model) gather and aggregate history cut the average model round‑trips per workflow from ~8.7 to ~5.5 — roughly a third fewer turns. Fewer turns means less waiting on generation and less back‑and‑forth to finish a task. Cheaper. The same change cut usage‑based cost ~24% — and, more importantly, held it flat as the logbook grew to hundreds of entries, because scripts carry the history the model used to re‑read every run. That's the quiet lesson: the reliability work I did for correctness turned out to be the same work that made the agent quicker and less expensive. Determinism isn't a tax on speed — here it bought all three. The work: three repositories, three runtimes Everything below the core — runtime, data access, governance, file layout — is where the effort went. 1 · Mainline — Copilot CLI + MCP. The upstream, most feature‑complete build. Runs as a primary agent in the GitHub Copilot CLI on Claude Opus 4.8; data services are discovered through MCP. It carries the heaviest governance: a Spec Kit layer (spec‑driven‑development agents, a constitution + templates, and 50+ per‑feature spec artifacts gated at PR time) plus an add‑on framework. The richest architecture — and the most complex to operate. 2 · Scout‑native. A thin wrapper loads the exact same core onto Microsoft Scout — again on Claude Opus 4.8 — but data access is re‑platformed onto Scout's native tooling instead of MCP subprocesses. No broker to configure; native tools negotiate their own auth. It adds two things the CLI can't do as cleanly: ✅ Scheduled automations — my morning briefing fires automatically on weekday mornings ✅ Cross‑session memory in place of hand‑off files The deterministic finalize gate stays fully intact. 3 · Foundry Local — on‑device NPU. The genuine outlier and the most involved port: a Python re‑implementation that runs the model — qwen2.5‑7b, an open ~7‑billion‑parameter model — 100% locally on the device's NPU (a Snapdragon X Elite Copilot+ PC) via Foundry Local's OpenAI‑compatible server. The agent loop, an MCP client, skill loading, and a distinct finalize pipeline all had to be rebuilt outside the CLI. The model never leaves the machine; only data connectors reach out when connected. The trade‑offs are real — modest throughput and a fixed context window — but so is the payoff: offline, private, near‑zero marginal cost. The effort This wasn't a weekend spike. Across the three code bases (plus a clean isolation clone I kept as an A/B baseline): ~340–380 commits per repository, three versions maintained in parallel 17 skills in each cloud build; 18 in the Foundry port ~37 scripts in the streamlined Scout build, up to ~97 in the governed Mainline build A Spec Kit governance layer with 50+ feature specs on Mainline A four‑part cost study and two written deliverables: an architecture whitepaper and a 20‑slide comparison deck The analysis and reporting The whitepaper and deck do two jobs. First, they document each runtime as a layered diagram — runtime, core, skills, scripting/validation, external services — so the differences are visible at a glance. Second, they convert the architecture fork into economics: a study that measured the actual token footprints of each repo and priced runs across billing models and hardware. The four dimensions: per‑skill cost, optimized‑vs‑out‑of‑the‑box, Copilot CLI vs Scout, and cloud vs local NPU. By the numbers The study priced measured token footprints at frontier‑model rates (treat the dollars as ±30% — the relative conclusions are far more robust than the absolute figures): Per skill: roughly $0.6–$1.4 per run usage‑based — or a single flat "premium request" under request‑based billing The determinism dividend: optimized, script‑driven skills cut model round‑trips ~8.7 → ~5.5 and usage‑based cost ~24% — and held cost flat as the logbook grew Scout vs CLI: Scout ran ~37% cheaper across a five‑command session and consumed none of the premium‑request allowance Cloud vs local: on‑device NPU inference came in 50–3,400× cheaper in cash than cloud — at the cost of throughput, context, and first‑pass reliability A full active day (~4 runs) landed around a few dollars usage‑based The headline isn't any single figure — it's the shape: cloud cents buy first‑pass reliability, on‑device near‑zero cost trades your time, and determinism makes either one cheaper and steadier. What held up The core is portable. The same agent, skills, and validation gate ran under all three runtimes. Good separation of concerns paid off. Determinism pays three ways — faster, cheaper, and more reliable (detailed above). It was the highest‑leverage change I made. Managed cloud wins the day job. Scout is the best daily driver: reliability gate intact, lower setup friction, scheduling + memory, and cheaper across a multi‑command session because it caches the bootstrap. On‑device is strategic — but reliability is the tax. Local NPU inference is dramatically cheaper in cash. We ran an in‑depth test pass across every function and closed the gaps it surfaced — yet the smaller model that makes Foundry Local possible still hallucinates and drops instructions often enough on the first pass to matter. Each re‑run is nearly free in dollars, but it costs real time to catch and correct. The winning pattern is hybrid. Draft and triage locally for ~nothing; escalate the correctness‑critical steps to cloud Opus 4.8, paying only where it buys first‑pass reliability. Three runtimes, side by side Figure: Three runtimes, one shared core. Only the top rows — runtime, model, data access, and governance — differ; the behavioral core, skill library, validation gate, and outputs are identical across all three. Capability Mainline (Copilot CLI) Scout‑native Foundry Local (NPU) Runtime Copilot CLI (cloud) Scout (cloud, managed) On‑device NPU Model Claude Opus 4.8 Claude Opus 4.8 qwen2.5‑7b (open, ~7B) Data access MCP Native tools MCP via local client Governance Spec Kit + PR gate Behavioral rules Behavioral rules Scheduling + memory ❌ ✅ ❌ Runs fully offline ❌ ❌ ✅ Marginal cost / run cloud per‑token cloud per‑token (cheaper/session) ≈ free Best for Framework development Daily production Offline / privacy / bulk When to use each Daily CSA workflows → Scout‑native. Managed, cheaper across a session, reliable, and it doesn't burn your Copilot request allowance. Building or versioning the framework → Mainline. Spec Kit governance and the add‑on system earn their keep here. Offline, air‑gapped, or sensitive data → Foundry Local. 100% on‑device inference. Bulk / high‑volume / non‑critical → Foundry Local. Zero marginal cost. Must be right on the first pass → Cloud Opus 4.8. The cents are worth it. Mixed, cost‑sensitive workload → Hybrid. Local draft → cloud escalate. Closing Thoughts The most useful reframe from this work: the three architectures aren't competitors — they're a portfolio. A managed cloud daily‑driver (Scout), a governed development platform (Mainline), and a sovereign on‑device runtime (Foundry Local). The job is to match the runtime to the task, not to crown one winner. And the same lesson that applies to Azure infrastructure applies to agents: build reliability into the system, not into good intentions. Because CSA‑Sherpa keeps its guarantees in code, I could change the entire execution model underneath it — cloud CLI, managed cloud, on‑device NPU — and the agent still behaved the same way. That portability is the dividend of a deterministic design. These workflows are genuinely complex, and that's exactly where the small model shows its limits: even after closing the gaps our testing surfaced, it still hallucinates and drops instructions often enough on the first pass to be a real cost. That's the honest trade‑off — near‑zero dollars, paid back in review‑and‑retry time — and it's why my recommendation lands on hybrid: let the small model draft where it's cheap and low‑risk, and escalate anything that has to be right the first time to cloud Opus 4.8. I use the agent in Microsoft Scout daily, as part of my personal production process. I did use AI to help draft and format this post — fittingly, the very agent it describes. The architecture, the analysis, and the conclusions are my own. Thanks for reading.Introduction to Prompt Engineering
With GPT-3, GPT-3.5, and GPT-4 prompt-based models, the user interacts with the model by entering a text prompt, to which the model responds with a text completion. Basic concepts and elements of GPT prompts Prompt components Instructions Primary Content Examples Cue Supporting content Prompts Basics Text prompts are how users interact with GPT models GPT models attempt to produce the next series of words that are most likely to follow from the previous text. Prompts | Best Practices Be Specific: Leave as little to interpretation as possible. Restrict the operational space Be Descriptive: Use analogies Double Down:Sometimes you may need to repeat yourself to the model. Give instructions before and after your primary content, use an instruction and a cue, etc. Order Matters:The order in which you present information to the model may impact the output. Whether you put instructions before your content (“summarize the following…”) or after (“summarize the above…”) can make a difference in output. Even the order of few-shot examples can matter. This is referred to as recency bias. Give the model an “out” :It can sometimes be helpful to give the model an alternative path if it is unable to complete the assigned task. For example, when asking a question over a piece of text you might include something like "respond with ‘not found’ if the answer is not present". This can help the model avoid generating false responses Prompt components Instructions When we show up to the present moment with all of our senses, we invite the world to fill us with joy. The pains of the past are behind us. The future has yet to unfold. But the now is full of beauty simply waiting for our attention. Instructions are likely the most commonly used prompt component Instructions - instruct the model on what to do Space efficiency TABLES As shown in the examples in the previous section, GPT models can understand tabular formatted data quite easily. This can be a space efficient way to include data, rather than preceding every field with name (such as with JSON). WHITE SPACE Consecutive whitespaces are treated as separate tokens which can be an easy way to waste space. Spaces preceding a word, on the other hand, are typically treated as part of the same token as the word. Carefully watch your usage of whitespace and don’t use punctuation when a space alone will do. Advanced techniques in prompt design and prompt engineering Certain models expect a specialized prompt structure For Azure OpenAI GPT models, there are currently two distinct APIs where prompt engineering comes into play: Chat Completion API Completion API Each API requires input data to be formatted differently Use of affordances | Factual claims, Search queries and Snippets Factual claims: John Smith is married to Lucy Smith John and Lucy have five kids John works as a software engineer at Microsoft Search queries: John Smith married to Lucy Smith John Smith number of children John Smith software engineer Microsoft Snippets: [1] … John Smith’s wedding was on September 25, 2012 … [2] … John Smith was accompanied by his wife Lucy to a party [3]John was accompanied to the soccer game by his two daughters and three sons [4] … After spending 10 years at Microsoft, Smith founded his own startup, Tailspin Toys [5] John M is the town smith, and he married Fiona. They have a daughter named Lucy System message framework and template recommendations for Large Language Models (LLMs) Define the model’s profile, capabilities, and limitations for your scenari Define the specific task(s) Define how the model should complete the tasks Define the scope and limitations Define the posture and tone Define the model's output format Define the language and syntax Define any styling or formatting Provide example(s) to demonstrate the intended behavior of the mode Describe difficult use cases Show the potential “inner monologue” Define additional behavioral guardrail Identify and prioritize the harms you’d like to address.
