---
name: workspace-health-dashboard
description: 'Audit test health and typecheck status across the current multi-repo workspace and produce a consolidated dashboard.'
argument-hint: Scope or focus, for example "all repos", "web only", "backend only", "changed repos only", or "failures first"
agent: agent
---

Create a workspace-wide health dashboard for the current multi-repo workspace, focused on typecheck status, test health, and execution readiness.

## Scope

- Treat the current VS Code multi-root workspace as the audit boundary.
- Include all repositories unless the user narrows scope.
- Prefer documented repo-native commands.
- Stay read-only unless the user explicitly asks for remediation.

## Discovery

For each repository, identify:

- primary stack and languages
- available typecheck commands
- available test commands
- whether commands are fast, medium, or expensive
- whether the repository is dirty

## Execution Rules

- Run the smallest reliable command set that gives useful signal.
- Prefer repo-native commands such as `npm run typecheck`, `pnpm typecheck`, `npm test`, `pnpm test`, `pytest`, `xcodebuild test`, or documented equivalents.
- If a repo lacks a formal typecheck or test path, mark it clearly.
- Separate command failures from code quality findings.

## Output Requirements

Produce one consolidated markdown dashboard with these sections:

1. Executive summary
2. Workspace scorecard
3. Repo-by-repo status
4. Failure clusters
5. Recommended actions
6. Command log summary
7. Risks and blockers

## Required Tables

### Workspace scorecard

| Repo | Typecheck | Tests | Dirty | Health | Notes |

### Failure clusters

| Repo | Category | Signal | Likely cause | Suggested next step |

### Action table

| Priority | Repo | Action | Effort | Expected impact | Safe to automate |

## Classification Rules

Use clear labels such as:

- `passing`
- `failing`
- `missing`
- `blocked`
- `not run`

Use health levels such as:

- `critical`
- `high`
- `medium`
- `low`

## Drill-Down Expectations

For each repo, include:

- what commands were run
- whether typecheck passed
- whether tests passed
- whether failures are setup, environment, flaky, or real code issues
- the smallest next action

## Default Behavior

If no argument is provided:

- audit all repositories in the workspace
- prioritize fast typecheck and unit-test signals first
- produce a single dashboard in the response

## Optional Saved Artifact

If the user asks to save the dashboard, write a dated markdown file under `/Users/sd9235/code/mygh/learning_ai_common_plat/.github/reports/health/`.

## Final Response Style

- start with the highest-signal outcome
- keep the summary concise
- surface blockers early
- end with a short numbered list of next actions