Session M3-04: Custom agents & agent profiles¶
Module: Agentic Workflows Difficulty: Advanced Prerequisites: Sessions M1-01 to M1-03, M2-01 to M2-04, M3-03 Duration: 3 hours (1 hour trainer content + 2 hours lab)
Overview¶
Custom agents define narrow roles in Markdown that teams can review. Start with a weak test-writer profile that enables every tool and permits source edits. After its boundary test fails, replace it with a profile that may read source, edit tests, and run one known command. The profile must stop when production code needs to change.
Learning outcomes¶
- Explain the
.agent.mdprofile structure and current frontmatter. - Detect a boundary violation with a repeatable contract test.
- Use documented tool aliases as a least-privilege allowlist.
- Write a stop condition that hands production conflicts to a human.
- Load the profile in the CLI and check that a source-code conflict stops the run.
- Choose between instructions, custom agents, skills, and MCP.
- Inspect an LSP definition/reference lookup and distinguish a caller from an export.
- Record an owner and version marker for a shared profile.
Required access¶
You need Node.js 22.22.2 or later, Git, the standalone Copilot CLI with live custom-agent loading, and a working JavaScript LSP server. Stop if any required access is missing. A manual scenario does not replace the live run or lookup.
| Resource | Location |
|---|---|
| Trainer guide | trainer-content/README.md |
| Slides | slides.md |
| Lab | lab/README.md |
| Language-server setup | lab/lsp-setup.md |
| Weak profile | lab/starter/agent-project/.github/agents/weak-test-writer.agent.md |
| Tightened profile | lab/solution/agent-project/.github/agents/test-writer.agent.md |
| Profile reference | lab/solution/agent-project/PROFILE-REFERENCE.md |