Index of every page on this site: llms.txt
All skills
github avatar

/foundry-agentic-loop

@f10fdb8 official
by githubgithub/awesome-copilot41k stars
5,399
Run it yourself

Build agentic solutions on Microsoft Foundry, GitHub and Azure: hosted agents, toolbox tools and skills, Foundry IQ grounding and a React/FastAPI app. Three modes: rapid prototyping in an existing Foundry project; spec-driven development (SDD), where the foundry-agentic-loop-workflow runs Specify, Implement, Verify and Deploy to create the Foundry project and deploy with azd, or one phase runs alone; and Improve, which refines this skill and the workflow from past runs. Use for Agentic Loop requests and canvas actions, or end-to-end agent solutions from architecture to Azure deployment. Not for isolated Azure administration or unrelated UI edits.

  • 55 files
  • 352.6 KB
  • Updated last month
  • GitHub
Use this Skill: https://skilld.dev/gh/github/awesome-copilot/foundry-agentic-loop

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

referencesimplementation-contract.md

≈948 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Cross-component implementation contract

Reference for the foundry-agentic-loop skill: the small contract that must be frozen before frontend, backend, agent, ingestion, or infrastructure work proceeds in parallel. Copy this table into the root README.md (rapid) or the Implementation contract section of ./docs/spec.md (SDD) and replace every placeholder.

Contract table

Contract Frozen value Producer / authority Consumers Verification
azure.yaml service names <frontend>, <backend>, <agent>, <ingestion> as applicable Infrastructure azd hooks, deployment scripts, CI azd config list and manifest review
Environment variables Exact name, type, required/optional status, and default for each variable Provisioning output or owning service Every component that reads it Generated env map matches code/config reads
Hosted-agent protocol Responses API version and OpenAI-compatible /responses endpoint Foundry hosted agent Backend/client Contract test invokes deployed endpoint
Responses metadata Allowed keys and JSON value types; no implementation-specific objects Backend/client contract Agent and telemetry Schema validation
Toolbox endpoint <project-endpoint>/toolboxes/<name>/versions/<version>/mcp?api-version=v1; define pinned versus promoted-version behavior azd ai toolbox create / provisioning Agent MCP bridge MCP initialize and tools/list
MCP tools Exact tool names, input JSON schemas, output/error shapes Tool owners Agent Schema snapshot and discovery test
Model inference Endpoint, deployment name, model/version/SKU, and token scope (https://ai.azure.com/.default) Foundry provisioning Agent framework Authenticated smoke test
Response body Exact JSON shape, status/error mapping, and streaming behavior Backend API Frontend/client API contract test
Citations Source, section, anchor/chunk id, and claim association Foundry IQ/toolbox Agent, backend, frontend Known-answer citation test
Grounding refusal Exact deterministic refusal text used when no cited evidence supports the answer Product/API contract Agent, backend, evals No-supported-source test
Runtime versions Python, Node, hosted-agent runtime and each direct dependency, including github-copilot-sdk, resolved during implementation to the latest stable release, API-verified before generating code and pinned exactly; exact image tags (digest preferred), never a floating "latest" Official package index, the build's package index and registries All services and CI Lockfile/manifests match CI; the build's index or registry serves each pinned version

For environment variables, add one row per value when that is clearer than a comma-separated list. Record both the producer and every consumer; this catches values that provisioning emits under one name while runtime code expects another.

Reconciliation before merge

  1. Compare every azure.yaml service and hook target with the frozen service names.
  2. Search code, Bicep outputs, azd hooks, CI, and docs for each environment-variable producer and consumer.
  3. Exercise the Responses endpoint with contract-valid metadata and validate response/citation/refusal schemas.
  4. Initialize the toolbox endpoint, compare discovered MCP names and schemas with the frozen snapshot, and fail on drift.
  5. Confirm model endpoint, deployment name, and token scope agree across provisioning and agent configuration.
  6. Compare runtime and package manifests with the resolved version row and run compatibility checks in the generated application.
  7. Update the contract and all consumers in one change when an intentional interface change is required.

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at f10fdb8. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated last month

README badge

README badge for github/awesome-copilot/foundry-agentic-loop