microsoft--agent-framework/dotnet/samples/02-agents/Harness/BuildYourOwnClaw/Claw_Step03_ScalingCapabilities

Scaling its capabilities (Post 3) — .NET

The third runnable sample from the "Build your own claw and agent harness with Microsoft Agent Framework" blog series (Part 3 — Scaling its capabilities). It builds on Post 2's personal finance assistant and makes it more capable along four axes.

What this sample demonstrates

• Skills — finance know-how (valuation, risk-scoring) is packaged as discoverable SKILL.md files under skills/, which the agent loads on demand. The sample builds its own provider with AgentSkillsProviderBuilder.UseFileSkills([skillsDir], scriptRunner: new SubprocessScriptRunner().RunAsync) so the skills' Python scripts can run, and sets DisableAgentSkillsProvider = true to replace the harness default. Optionally folds in centrally-managed Foundry skills discovered live from a Foundry Toolbox MCP endpoint via FoundrySkills.ConnectAsync(...) + UseMcpSkills(...) (opt-in; see below).
• Shell — a LocalShellExecutor confined to the trade-confirmation vault (working/confirmations/) lets the agent tidy the accumulated confirmation files (reorganize into year/month, rename to YYYY-MM-DD_TICKER_BUY|SELL.txt). ConfineWorkingDirectory re-anchors every command to the vault and a ShellPolicy deny-list pre-filters obviously destructive commands. Exposed as the run_shell tool, which prompts for approval before each command runs. (The deny-list is a UX guardrail, not a security boundary — for hard isolation use a DockerShellExecutor.)
• CodeAct — a HyperlightCodeActProvider gives the agent a sandboxed Python interpreter to crunch portfolio numbers by writing and running code. It runs on Hyperlight (a micro-VM), so it requires hardware virtualization. The guest module path is resolved automatically from the Hyperlight.HyperlightSandbox.Guest.Python NuGet package via PythonGuestModule.GetModulePath().
• Background agents — a lean, web-search-only ResearchAgent is registered via HarnessAgentOptions.BackgroundAgents, exposing the background_agents_* tools so the main agent can fan out per-ticker research concurrently and aggregate the findings.

Prerequisites

1. A Microsoft Foundry project with a deployed model (e.g. gpt-5.4).
2. Azure CLI installed and authenticated (az login).
3. (For CodeAct) a host with hardware virtualization enabled (Hyperlight runs the Python interpreter in a micro-VM).

Environment variables

export FOUNDRY_PROJECT_ENDPOINT="https://your-project.services.ai.azure.com/api/projects/your-project"
# Optional (defaults to gpt-5.4)
export FOUNDRY_MODEL="gpt-5.4"

# Optional — enable centrally-managed Foundry skills (Foundry Toolbox MCP endpoint URL):
export FOUNDRY_TOOLBOX_MCP_SERVER_URL="https://your-project.services.ai.azure.com/.../toolboxes/your-toolbox/mcp?api-version=v1"

When FOUNDRY_TOOLBOX_MCP_SERVER_URL is not set, the sample runs with the local file skills only and prints a note.

Running

cd dotnet
dotnet run --project samples/02-agents/Harness/BuildYourOwnClaw/Claw_Step03_ScalingCapabilities

What to expect

The sample starts an interactive loop in execute mode (quick lookups don't need a plan). Try these in order:

1. Value MSFT for me. — the agent loads the valuation skill and follows its instructions (reading references and running its script).
2. Score the risk of my portfolio. — the agent reads portfolio.csv and loads the risk-scoring skill.
3. /mode plan, then Tidy up my trade confirmations. — switching to plan mode first makes the agent inspect working/confirmations/ and propose a reorganization plan before touching anything; once you approve it switches to execute and uses the shell to reorganize and rename the files, prompting you to approve each command.
4. Work out the total value of my portfolio. — the agent writes and runs Python via CodeAct.
5. Research MSFT, NVDA and SPY and summarize the latest news. — the agent fans the tickers out to the background research agent and aggregates the results.
6. What's the capital of France? — with a financial-agent-rules skill published to your Foundry toolbox and Foundry skills enabled (FOUNDRY_TOOLBOX_MCP_SERVER_URL), the agent loads it, recognizes the question is off-topic, and politely declines, steering you back to finance.

See the Part 3 blog post for more on the financial-agent-rules skill — including the SKILL.md to publish to your Foundry toolbox.