Code quality dashboard with weighted scoring. Runs type checker, linter, test suite with coverage, and dead code detection in parallel, then computes a 0-10 quality score. Tracks trends over time. Use when user says "health check", "code quality", "project health", "quality score", "how healthy is this project", or wants a quick quality snapshot. Do NOT use for code review (use /review), running tests alone (use /test), or security audit (use /audit).

Report-only quality dashboard that computes a weighted score from multiple quality signals. Never auto-fixes anything.

Arguments

No arguments: run the full health check on the current project.
--no-trend: skip trend tracking, report only.

Process

1. Detect Project Tools

Read the project's package manager config, lockfile, and scripts to identify:

Signal	Detection method
Type checker	`tsc`, `mypy`, `pyright`, `go vet` in scripts or devDependencies
Linter	`eslint`, `ruff`, `golangci-lint`, `clippy` in scripts or devDependencies
Test runner	`jest`, `vitest`, `pytest`, `go test`, `cargo test` in scripts or devDependencies
Coverage tool	`--coverage` flag, `c8`, `nyc`, `coverage.py` in scripts or devDependencies
Formatter	`prettier`, `black`, `gofmt`, `rustfmt` in scripts or devDependencies

If a tool is not present, skip that signal and redistribute its weight proportionally across the remaining signals.

2. Run Checks in Parallel

Execute all detected tools simultaneously:

Type checker: run with no-emit flag. Count errors.
Linter: run in report mode. Count errors and warnings separately.
Test suite: run with coverage enabled. Capture pass/fail counts and coverage percentages for statements, branches, functions, and lines.
Dead code detection: grep for exported symbols and check if they are imported elsewhere. Count unused exports.

3. Compute Scores

Each signal produces a 0-10 sub-score:

Signal	Weight	Scoring
Tests	30%	10 if all pass, subtract 1 per failure, floor at 0
Coverage	25%	coverage percentage divided by 10, e.g. 95% = 9.5
Lint	20%	10 minus 0.5 per warning minus 1 per error, floor at 0
Types	15%	10 if zero errors, subtract 1 per error, floor at 0
Dead code	10%	10 minus 0.5 per unused export, floor at 0

Composite score: weighted sum of sub-scores, rounded to one decimal.

4. Report

Present results as a summary table:

Signal	Raw Result	Score	Weight	Weighted
Tests	142 pass, 0 fail	10.0	30%	3.00
Coverage	87.3% avg	8.7	25%	2.18
Lint	2 warnings, 0 errors	9.0	20%	1.80
Types	0 errors	10.0	15%	1.50
Dead code	3 unused exports	8.5	10%	0.85
Total				9.3

Include a one-line assessment:

Score range	Assessment
9.0 - 10.0	Healthy. No action needed.
7.0 - 8.9	Acceptable. Minor issues to address.
5.0 - 6.9	Degraded. Schedule cleanup.
0.0 - 4.9	Critical. Prioritize remediation.

5. Track Trends

Unless --no-trend is passed:

Identify the project by its git remote URL or directory name.

Append a JSON line to ~/.claude/telemetry/health-<project>.jsonl:

{"timestamp":"2026-04-04T12:00:00Z","score":9.3,"tests":{"pass":142,"fail":0},"coverage":87.3,"lint":{"warnings":2,"errors":0},"types":{"errors":0},"deadCode":{"unused":3}}

If previous entries exist, compare the current score against the last entry. Report the delta: "+0.3 since last check" or "-1.2 since last check".
Create the ~/.claude/telemetry/ directory if it does not exist.

Rules

Report only. Never modify source files, configs, or dependencies.
Show raw tool output excerpts when a sub-score is below 7.0 so the user can see what needs attention.
When a tool is missing, state which signal was skipped and note that the score reflects fewer dimensions.
All timestamps in GMT.
Trend files are append-only. Never truncate or rewrite them.

Related Skills

/review -- Detailed code review for specific changes.
/test -- Run the test suite with full output.
/audit -- Security-focused audit.