# Custom Copilot agents for test-automation teams

- Period: Mar 2026 – Sep 2026
- Domain: Enterprise test-automation teams (Quality engineering, Developer platforms and tooling)
- Role: Wrote the agents and skills, ran the enablement, published the catalog
- Key result: 27, agents and skills catalogued across seven SDLC phases
- Capabilities: AI adoption at enterprise scale; Agent architecture and orchestration
- Stack: GitHub Copilot agents, Agent skills, Playwright, Pact, Docs portal
- URL: https://farhanbaluch.com/work/copilot-agent-program

Custom agents and skills that generate UI, API and contract tests in each team's own conventions, rolled out with hands-on sessions and a searchable catalog of 27 agents.

**AI angle.** Teams had AI assistants, but generic prompts wrote generic tests that ignored each team's frameworks, page objects and naming, so adoption stayed shallow. Encoding those conventions into agents and skills is what made AI-generated tests mergeable, and the catalog is what turned individual agents into a toolkit the whole programme reuses.

**What this shows:**
- I write production agents and skills, not just prompts, tuned to how a team actually works.
- I know how agent platforms load and run skills, and design rollouts around it.
- I drive adoption: enablement sessions, a catalog, and a path from one team to many.

### Problem

Test-automation teams had access to AI assistants, but the tests they produced ignored each team's frameworks, page objects and naming conventions. Engineers rewrote most of the output, and adoption stayed shallow.

### Approach

Put each team's conventions into the agent, not into every prompt. Start with the highest-volume work (UI test generation), move down the test pyramid, and make the agents easy to find so teams reuse them instead of writing their own.

### What I built

- Custom agents and skills that generate Playwright tests in each team's own conventions, with hands-on sessions so teams actually used them.
- Agent workflows for Pact contract tests and API suites, so service changes are caught before end-to-end runs.
- A rollout plan for code-review skills, rewritten after finding that skills only load from a repository's default branch.
- A searchable catalog of 27 agents and skills, each mapped to one of seven SDLC phases, published in a docs portal.

### Outcome

- Engineers look for an existing agent before writing a new one, and the programme can see which SDLC phases still have no agent support.
- These agents became the packages that the [distribution platform](/work/agent-distribution) versions and installs.