Health Check

Comprehensive project health check that auto-detects relevant auditors, runs them in parallel, and produces a single prioritized report.

What It Does

• Scans project structure to detect which frameworks and patterns are used
• Launches all relevant auditors in parallel (6 always-run + conditional)
• Deduplicates findings that appear across multiple auditors (same file:line)
• Produces unified report with executive summary + grouped details
• Writes full results to scratch/health-check-*.md files

Usage

/axiom:health-check

Or via the audit command:

/axiom:audit all

Excluding Auditors

/axiom:health-check skip spritekit skip camera

Scoping to a Branch's Diff

For PR review, scope the audit to files changed on the current branch — auditors only inspect the files your PR touched, so findings stay relevant and don't pull in pre-existing issues from elsewhere in the repo.

/axiom:health-check diff

This compares the current branch against the merge-base with origin/main (falling back to origin/master, then main, then master if origin/main doesn't exist).

The report's first line declares the scope (changed files vs <base> (merge-base <SHA>, N files)) and lists the audited files. Findings outside that file list are dropped, even if an auditor surfaces them — the scope is a hard boundary.

If there are no changed Swift files vs the base, the command exits cleanly without launching any auditors.

Emphasis (Freeform)

Anything you write that isn't diff or skip <auditor> is treated as emphasis — the agent uses it to prioritize the executive summary and the order of the summary table, but it never excludes auditors. Useful when you want a full audit but care most about a specific area.

/axiom:health-check focus on memory leaks
/axiom:health-check I'm worried about Core Data migrations
/axiom:health-check prioritize accessibility before release

Combining Modifiers

All three modifiers compose freely:

/axiom:health-check diff skip camera focus on concurrency

This runs a diff-scoped audit, omits the camera auditor, and surfaces concurrency findings first in the report.

Auto-Detection

The agent detects which auditors are relevant by scanning for framework imports, file types, and code patterns. Six auditors always run (memory, security, accessibility, swift-performance, modernization, codable). Others activate based on what's found in your project.

Output

Results appear in two places:

1. Conversation — Executive summary with top 5 critical findings + summary table
2. scratch/ directory — Full detailed reports per auditor domain

Related

• Audit Command — Run individual auditors by domain
• Memory Auditor — One of the 26 individual auditors health-check orchestrates