| name | confidence |
| description | Analyze test suite quality using the Confidence CLI tool. Use when asked to assess test quality, review test health, generate a test confidence report, or when phrases like "run confidence", "test quality report", "how are the tests", "scan tests", "test health" are used. Also use proactively after completing major test-related work to validate quality. |
| allowed-tools | Bash(confidence *) |
Confidence โ Test Suite Quality Analysis
Analyze test suite quality through static analysis and git history using the confidence CLI.
Install
If confidence is not on PATH, build it from the repo:
go build -o confidence ./cmd/confidence
Or install globally:
go install github.com/timusus/test-confidence/cmd/confidence@latest
Commands
Default to the HTML report โ it's the most useful output for understanding a codebase. Only use terminal or JSON for quick checks or CI.
HTML report (preferred โ visual, self-contained)
confidence scan <path> --html > /tmp/confidence-report.html && open /tmp/confidence-report.html
HTML with historical trends (recommended for first run)
confidence scan <path> --html --trend-months 12 > /tmp/confidence-report.html && open /tmp/confidence-report.html
Terminal report (quick check)
confidence scan <path>
JSON output (for CI or programmatic use)
confidence scan <path> --json
With coverage data (optional, enriches analysis)
confidence scan <path> --coverage build/reports/kover/report.xml
confidence scan <path> --platform ios --coverage coverage.json
iOS projects (need platform flag)
confidence scan <path> --platform ios
Verbose per-file detail
confidence scan <path> --verbose
Compare to baseline
confidence scan <path> --since 3months
confidence scan <path> --compare-ref v1.0
confidence scan <path> --compare old.json
When to use
- User asks about test quality, test health, or test confidence for a project
- After completing a major feature or test refactor, to check impact
- When reviewing a project's test strategy or test suite
- When comparing test quality across projects or over time
- When asked to identify the highest-risk untested code
Interpreting results
The tool presents signals, not opinions. Focus on:
- Behavioral % โ proportion of tests checking outcomes vs implementation
- Where To Start โ prioritized recommendations ranked by impact
- Evidence โ observed maintenance cost from git history (how often refactors break tests)
- Risk Hotspots โ files with high churn AND fragile tests
- Coverage Gaps โ untested surfaces and business logic
- Patterns โ novel cross-file signals (mock blast radius, test freshness, production complexity shadow)
Notes
- The tool reads source files and git history only โ never compiles or executes tests
- All analysis is offline and deterministic
- Coverage data is optional but enriches every signal when available
- Works on Android/Kotlin and iOS/Swift codebases