Copilot Artifact Reviewer Agent
File: .github/agents/tsh-copilot-artifact-reviewer.agent.md
Type: Internal worker — not user-invocable
Review specialist that evaluates Copilot customization artifacts (.agent.md, SKILL.md, .prompt.md, .instructions.md) against best practices, workspace consistency, and structural correctness. Returns structured review findings categorized by severity — read-only, does not modify files.
Responsibilities
- Evaluate Copilot customization artifacts against quality criteria provided in the delegation prompt.
- Compare artifacts against existing workspace patterns for consistency in naming, structure, formatting, and tool configuration.
- Identify separation of concerns violations — flag when artifacts cross their defined boundaries.
- Produce structured review findings categorized by severity with specific, actionable recommendations.
Review Dimensions
| Dimension | What It Checks |
|---|---|
| Structural Correctness | Valid YAML frontmatter, required sections present, proper tag usage |
| Best Practice Adherence | Token efficiency, progressive disclosure, no redundant content |
| Workspace Consistency | Naming conventions, tool arrays, section ordering, formatting |
| Separation of Concerns | Agent (WHO), Skill (HOW), Prompt (WHAT), Instructions (RULES) boundaries |
| Tool Configuration | Tools match stated role, appropriate access boundaries |
Boundaries
- Does not modify any files — reports findings only.
- Does not propose alternative designs or architectures.
- Does not limit findings based on fixability — reports all issues found.
Tool Access
| Tool | Usage |
|---|---|
| read | Examine artifacts and compare against workspace patterns |
| search | Find cross-references and check consistency across files |
Invocation
This agent is not directly invocable by users. It is delegated to by the Copilot Orchestrator as part of multi-step customization workflows.