Skip to content

Latest commit

 

History

History
42 lines (28 loc) · 2.24 KB

CLAUDE.md

File metadata and controls

42 lines (28 loc) · 2.24 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

What This Repo Is

A curated collection of custom agent profiles for Katalon Studio's StudioAssist feature. Each agent provides specialized AI behavior for different testing workflows (BDD, API testing, code review, test data generation). This is a template/reference repository — there are no build, test, or lint commands.

Requires Katalon Studio v10.3.2+ (v11.0.0+ for full agent profile support).

Agent Architecture

Each agent lives in Agents/<AgentName>/ and consists of exactly four files:

File Purpose
prompt.md System instructions defining the agent's role, workflow, output format, and constraints
profile.json Metadata: description (shown in StudioAssist UI) and enabled boolean
agent.json Tool permissions: maxToolCalls, autoToolApproval, per-server approvedTools/excludedTools
mcp.json External MCP server connections (empty "servers": {} means built-in servers only)

agent.json Tool Approval Behavior

  • approvedTools: Pre-approved, no user prompt needed
  • Not in either list: User prompted at runtime
  • excludedTools: Never available
  • All template agents use autoToolApproval: false

Built-in MCP Servers (always available, no mcp.json config needed)

  • katalon-studio — Read/modify test cases, keywords, objects, suites, BDD files, test data
  • katalon — Search/fetch Katalon documentation
  • katalon-testops — Manage test artifacts, requirements, sprints (v10.4.0+)

Note: Custom agents do not yet support references (bundled reference files). Do not add references/ directories or instruct prompts to read from them.

Conventions

  • Agent folder names: Alphanumeric, hyphens, underscores only (e.g., BDD-Agent, regression_testing_1)
  • Prompt structure: Role definition → Workflow steps → Output format → Constraints. Always reference Katalon conventions and built-in keywords.
  • Tool approval strategy: Pre-approve read/search/list operations; leave create/update operations to prompt the user.
  • Server+tool name limit: 64 characters max for {server_name}___{tool_name} identifiers (AI provider constraint).