Manual Agent Test Harness

The tests/manual/ directory contains 10 scenarios that test how AI agents interact with skern — discovery, command chaining, JSON parsing, error handling, and dedup reasoning.

Run these before releases to verify agent compatibility.

Setup

# 1. Set up isolated test environments in /tmp
make test-manual-setup

Running Scenarios

# 2. Test each scenario with your AI agent
cd /tmp/skern-manual-tests/01-fresh-project
cat PROMPT.md      # Read what to ask the agent
cat EXPECTED.md    # Read the pass criteria
# Open your AI agent and run the prompt
# Repeat for each scenario (01 through 10)

Each scenario directory contains:

File	Purpose
AGENTS.md	Agent instructions
PROMPT.md	What to ask the agent
EXPECTED.md	Pass/fail checklist

Generating Reports

# 3. Generate a markdown report (interactive pass/fail checklist)
make test-manual-report

Cleanup

# 4. Clean up temp dirs and platform markers
make test-manual-teardown

Scenario Coverage

The 10 scenarios cover:

1. Fresh project discovery
2. Skill creation workflows
3. Cross-platform installation
4. JSON output parsing
5. Overlap detection handling
6. Validation error recovery
7. Search and recommendation
8. Multi-skill management
9. Error handling and edge cases
10. Deduplication reasoning

See tests/manual/README.md for full details.