Skip to main content
AI/MLjmagly

uat-generate

Discover MCP tools and generate a phased UAT plan with agent-executable test specifications

Stars
141
Source
jmagly/aiwg
Updated
2026-05-31
Slug
jmagly--aiwg--uat-generate
View on GitHubRaw SKILL.md

// install — copy + paste into any project

mkdir -p .claude/skills && curl -fsSL https://raw.githubusercontent.com/jmagly/aiwg/HEAD/agentic/code/addons/uat-mcp/skills/uat-generate/SKILL.md -o .claude/skills/uat-generate.md

Drops the SKILL.md into .claude/skills/uat-generate.md. Works with Claude Code, Cursor, and any agent that loads SKILL.md files from .claude/skills/.

UAT Generate

Generate a phased User Acceptance Test plan by discovering tools from connected MCP servers and producing agent-executable test specifications.

Usage

# Generate UAT plan from all connected MCP servers (default: --mode mcp)
/uat-generate

# Explicit MCP mode
/uat-generate --mode mcp

# Target specific server
/uat-generate --mode mcp --server gitea

# Quick smoke test plan
/uat-generate --execution-mode quick

# Full comprehensive plan
/uat-generate --execution-mode full

# Custom output path
/uat-generate --output .aiwg/testing/uat/my-plan.md

# Interactive — ask questions about scope and priorities
/uat-generate --interactive

Parameters

--mode (default: mcp)

Test mode. Currently only mcp is supported. Future modes: api, ui.

--server (optional)

Target a specific MCP server by name. If omitted, discovers tools from all connected servers.

--output (optional)

Output path for the generated plan. Default: .aiwg/testing/uat/plan-{server}-{timestamp}.md

--execution-mode (default: standard)

Controls the breadth of generated tests:

Mode Description
quick Preflight + 1 happy path per tool. Minimal, fast.
standard Happy paths + key edge cases + negative tests. Balanced.
full All paths including exhaustive negative tests and E2E chains. Comprehensive.

--interactive (optional)

Prompt for scope decisions before generating:

  • Which tool categories to include/exclude
  • Priority ordering of phases
  • Whether to generate negative tests
  • Custom test data values

Workflow

Step 1: Discover MCP Tools

Enumerate all tools from connected MCP servers:

Discovering MCP tools...
  Server: gitea (mcp__gitea__*)
    Found 78 tools
  Server: filesystem (mcp__fs__*)
    Found 12 tools

Total: 90 tools across 2 servers

For each tool, capture:

  • Tool name and description
  • Parameter schema (required/optional, types, defaults)
  • Return type description
  • Any documented constraints or side effects

Step 2: Categorize Tools

Group tools into logical categories:

Categories:
  Repository Management: 15 tools (create, get, list, update, delete, fork, ...)
  Issue Tracking: 12 tools (create, get, list, edit, comment, close, ...)
  User & Org: 8 tools (get_user, search_users, get_orgs, ...)
  Actions/CI: 14 tools (list_runs, get_job, dispatch, ...)
  Labels & Milestones: 9 tools (create_label, edit_label, ...)
  Wiki: 5 tools (create, get, update, delete, list)
  Release & Tags: 8 tools (create_release, list_tags, ...)
  File Operations: 5 tools (get_file, create_file, update_file, ...)
  Misc: 2 tools (get_version, get_user_info)

Step 3: Design Phases

Apply the standard phase structure:

  1. Phase 0: Preflight — Connectivity, auth, server version
  2. Phase 1: Seed Data — Create test entities needed by later phases
  3. Phases 2-N: Per-Category — One phase per tool category
  4. Phase N+1: E2E Chains — Cross-category workflows
  5. Phase N+2: Cleanup — Delete all seeded data

Step 4: Generate Test Specs

For each tool in each phase, generate test cases following the uat-test-case.md template:

  • Happy path: Standard usage with valid inputs
  • Edge cases: Boundary values, empty strings, max lengths (standard and full modes)
  • Negative tests: Missing required params, invalid types, unauthorized access (standard and full modes)
  • E2E chains: Multi-tool workflows (full mode only)

Step 5: Wire Variables

Connect phases via stored variables:

Phase 1 (Seed): create_repo → Store: TEST_REPO_NAME
Phase 3 (Issues): create_issue(repo: ${TEST_REPO_NAME}) → Store: ISSUE_INDEX
Phase 4 (Comments): create_issue_comment(index: ${ISSUE_INDEX})
Phase N (Cleanup): delete_repo(name: ${TEST_REPO_NAME})

Step 6: Output Plan

Write the complete UAT plan to the output path. Display summary:

UAT Plan Generated
  Server: gitea
  Tools: 78
  Phases: 12
  Test cases: 165
  Negative tests: 32 (isolated)
  Estimated duration: ~20 minutes
  Output: .aiwg/testing/uat/plan-gitea-20260227.md

Review the plan, then execute with:
  /uat-execute .aiwg/testing/uat/plan-gitea-20260227.md

Plan Output Format

The generated plan follows the uat-phase.md template. See templates/uat-phase.md for the full format.

Error Handling

No MCP Servers Connected

No MCP servers detected.

To use UAT-MCP, you need at least one MCP server connection.
Check your MCP configuration and try again.

Server Discovery Fails

Failed to discover tools from server: {name}
Error: {details}

Continuing with remaining servers...

Zero Tools Found

No tools discovered from {server_name}.

The server is connected but exposes no tools.
This may indicate a configuration issue.

References

  • Agent: @$AIWG_ROOT/agentic/code/addons/uat-mcp/agents/uat-planner.md
  • Template: @$AIWG_ROOT/agentic/code/addons/uat-mcp/templates/uat-phase.md
  • Schema: @$AIWG_ROOT/agentic/code/addons/uat-mcp/schemas/uat-plan.yaml