agentleFS
Sign inStart free
stevemarksavastevemarksava/superteam/skills/azure-ai-foundry-dev/SKILL.md

azure-ai-foundry-dev

Skill1 starsChanged 2 months agoMITView on GitHub

Build AI agents on Azure AI Foundry — project setup, model choice, SDK-based agents, threads, tools, observability.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    AI Foundry project setup: Create a Foundry project (hub + project resource pair).

    line 18
  2. Very specific:

    Create agent via SDK: Use the azure-ai-projects SDK (Python) or @azure/ai-projects (TypeScript).

    line 20
  3. Distinctive:

    Every function tool has a validated JSON schema — no free-form tool input

    line 34

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/azure-ai-foundry-dev — Agent Development on Azure AI Foundry

Azure AI Foundry is Microsoft's managed platform for building, evaluating, and running AI agents. This skill takes a project from zero to a running agent with tools wired in and telemetry flowing.

When to invoke

  • Building an agent that must run under Azure governance (Entra, Defender, Purview)
  • A Copilot Studio agent needs a custom backend model or complex tool orchestration
  • Nexus + Bob are standing up an enterprise AI service on the Microsoft platform
  • Diesel needs backend Python for agent tool implementations

Steps

  1. AI Foundry project setup. Create a Foundry project (hub + project resource pair). Bind to a Log Analytics workspace for telemetry. Assign role-based access via Entra groups, not users. Note the project connection string.
  2. Model choice. Verify the current Foundry model catalogue at author time — model IDs and GA status shift monthly. Typical choices as of authoring: gpt-4o / gpt-4o-mini for general reasoning, mai-code-1.1-flash for coding tasks (Microsoft in-house), phi-4 for lightweight reasoning. Athanasios confirms latest IDs and regional availability before deployment.
  3. Create agent via SDK. Use the azure-ai-projects SDK (Python) or @azure/ai-projects (TypeScript). Verify current SDK version at author time. Define the agent with: name, model deployment, instructions, tools, top-level metadata.
  4. Thread management. Threads are conversation state. Create per user session. Persist thread IDs mapped to your app's user identity. Truncate or summarise on token pressure — do not silently drop context.
  5. Tool wiring. Attach tools: Code Interpreter (Python sandbox), File Search (vector-indexed docs), Function Calling (your REST endpoints), Bing Grounding, or MCP-based tools. Each function tool has a JSON schema; validate inputs before execution.
  6. Observability. Enable Foundry's built-in tracing → Application Insights. Log agent runs, tool calls, token usage, latency. Wire alerts on error rate and cost overruns. This telemetry is what G reads for cost, and what Coach reads for pattern detection.

Deliverable format

  • agent.py (or .ts) — agent definition, tools, and a runnable smoke test
  • foundry-config.md — project name, model deployments, connection string source (Key Vault reference)
  • tools/*.json — one JSON schema per function tool
  • Application Insights dashboard link with cost + latency panels

Non-negotiables

  • Model IDs verified against the live Foundry catalogue at author time — no version from memory
  • Connection strings and API keys via Key Vault + Managed Identity — never inline
  • Every function tool has a validated JSON schema — no free-form tool input
  • Tracing is on from day one — you cannot debug what you did not log

References

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free