feat: add /speckit.manual-test command for generating manual test guides

[Vishuvishu]

Summary

Adds a new /speckit.manual-test command that generates a comprehensive manual test guide (test.md) for any feature developed with spec-kit. This fills a gap in the SDD workflow: after /speckit.plan produces contracts and acceptance scenarios, there is no built-in way to generate human-executable (curl-based) test documentation.

Files changed

File	Change
templates/commands/manual-test.md	New command template
templates/manual-test-template.md	Scaffold template installed into .specify/templates/
templates/commands/plan.md	Added manual-test handoff + optional next-steps block in Completion Report

Position in the SDD workflow

/speckit.specify → /speckit.clarify (opt) → /speckit.plan
  → /speckit.manual-test  ← NEW (optional quality gate)
  → /speckit.checklist    (opt)
→ /speckit.tasks → /speckit.analyze (opt) → /speckit.implement

Runs after plan (needs contracts + acceptance criteria to exist) and before tasks (the test guide informs implementation). Same optional-quality-gate position as /speckit.checklist.

What the command generates

A test.md file in the feature directory containing:

• Prerequisites block — service URL, shell variable setup for tokens and IDs
• One section per resource group — e.g. Provider Management, Verifications, Transient Flows
• TEST-NN numbered test cases each with:
  • A **What it tests** line tracing back to a specific FR-### or acceptance scenario
  • Happy-path curl command with assertions
  • At least one ❌ Bad case block (wrong auth, invalid input, wrong state, etc.)
• Edge Cases & Security Tests — 401/403 per role, PII log inspection, field boundary validation
• State machine tests — valid transitions, invalid transitions, lazy expiry detection
• Full end-to-end smoke test — single copy-paste bash script with step separators
• Quick reference table — one row per test, method, path, auth, expected codes

Changes to plan.md

Added speckit.manual-test to the handoffs frontmatter so agents that support handoff menus surface it automatically. Added an explicit optional-next-steps block in the Completion Report so the AI always mentions the command after plan finishes — users do not need to know it exists.

Testing

The command follows the same structure as existing commands (checklist.md, analyze.md): pre-execution hook check, artifact loading via check-prerequisites.sh, generation, post-execution hook check, completion report. No changes to CLI infrastructure are required — the command is picked up automatically by the existing command-discovery logic in base.py (templates/commands/ scan).

Checklist

• New command template added to templates/commands/
• New scaffold template added to templates/
• plan.md handoffs and Completion Report updated
• Hook keys use consistent naming (before_manual_test / after_manual_test)
• Follows existing command structure (pre-hooks, outline, post-hooks, completion report, done-when checklist)
• No changes to CLI source code required