# Audit

> Run a comprehensive backend quality audit across all eight dimensions. Orchestrates scan, critique, harden, distill, verify, and humanize skills, deduplicates findings, and produces a unified report with verdict. Use when assessing overall codebase health. Triggers: 'audit backend', 'full quality check', 'run audit', or /axiom:audit. Do NOT use for targeted checks — use individual skills instead.

- Skill: `lvlup-sw/audit` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add lvlup-sw/audit`
- Raw SKILL.md: https://api.skillmd.com/api/skills/lvlup-sw/audit/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: lvlup-sw (https://skillmd.com/u/lvlup-sw)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/lvlup-sw/audit

---


# Audit — Comprehensive Backend Quality Assessment

## Overview

The anchor skill that orchestrates all other axiom skills to produce a comprehensive backend quality report. Runs deterministic checks and qualitative assessments across all 8 dimensions, deduplicates findings, computes per-dimension metrics, and delivers a unified verdict.

## Triggers

**Use when:**
- Assessing overall codebase or module health
- Preparing for a release or major refactor
- Onboarding to understand existing technical debt
- User says "audit", "full quality check", "assess quality"

**Do NOT use when:**
- Targeting a specific concern (use `axiom:critique`, `axiom:harden`, `axiom:distill`, or `axiom:verify` directly)
- Running only deterministic checks (use `axiom:scan`)
- Performing workflow-specific review (workflow tools orchestrate axiom, not the reverse)

## Process

### Step 1: Scope Resolution

Determine the assessment scope from the `scope` argument:
- **File:** Assess a single file
- **Directory:** Assess all files in a directory (recursive)
- **Codebase:** Assess the entire project (default: current working directory)

Exclude by default: `node_modules/`, `dist/`, `.git/`, binary files, generated files.

### Step 2: Deterministic Scan

Run `axiom:scan` with `dimensions: all` for the resolved scope. This produces the mechanical findings that ground the qualitative assessment.

### Step 3: Qualitative Assessment

Run each specialized skill in sequence, passing the scope:

1. **`axiom:critique`** — Architecture (DIM-6) + Topology (DIM-1)
2. **`axiom:harden`** — Observability (DIM-2) + Resilience (DIM-7)
3. **`axiom:distill`** — Hygiene (DIM-5) + Topology (DIM-1)
4. **`axiom:verify`** — Test Fidelity (DIM-4) + Contracts (DIM-3)
5. **`axiom:humanize`** — Prose Quality (DIM-8)

Each skill produces findings in the standard format: `@skills/backend-quality/references/findings-format.md`

### Step 4: Deduplicate and Aggregate

Merge findings from all skills using these rules:
- **Same evidence + same dimension:** Merge into single finding (keep most detailed explanation)
- **Same evidence + different dimensions:** Keep both (finding genuinely spans two concerns)
- **Same pattern + different files:** Keep as separate findings
- **Deterministic + qualitative for same issue:** Merge, mark as `deterministic: true`

### Step 5: Coverage Check

Verify all 8 dimensions were assessed. If any dimension has zero findings and zero checks:
- **Warning:** "DIM-N ({name}) was not assessed — no checks or findings produced"
- This may indicate the scope doesn't contain code relevant to that dimension

### Step 6: Compute Verdict

Apply the verdict exactly as defined in `@skills/backend-quality/references/scoring-model.md` (single source of truth). Do not redefine thresholds here; compute and report the inputs required by the model (HIGH/MEDIUM/LOW counts).

Compute per-dimension metrics:
- Pass rate (deterministic checks only)
- Finding count (all findings)
- Severity distribution (HIGH / MEDIUM / LOW)

### Step 7: Produce Report

Output the structured report per the template in `@skills/backend-quality/references/scoring-model.md`:

```markdown
# Backend Quality Report

**Scope:** [assessed scope]
**Verdict:** [CLEAN | NEEDS_ATTENTION]
**Date:** [current date]

## Summary
[Per-dimension table with findings, severity counts, pass rates]

## HIGH-Priority Findings
[Findings requiring immediate attention]

## MEDIUM-Priority Findings
[Findings to address soon]

## LOW-Priority Findings
[Polish and minor items]

## Dimensional Coverage
[Which dimensions assessed, any gaps]

## Recommendations
[Top 3-5 prioritized action items]
```

## Pairing Contract Check

When the assessed scope contains a project-specific invariants skill (any loaded skill with frontmatter `pairs-with: axiom:design`), audit runs an advisory check against the pairing contract documented at `@skills/backend-quality/references/pairing-contract.md`.

**Two checks, both LOW severity, both advisory only:**

1. **Overlap declaration.** The paired skill must declare `axiom_overlap: DIM-N` for at least one of its invariants. A paired skill with zero overlap declarations is technically compliant (axiom_overlap is optional per the contract) but suggests the pairing isn't specializing axiom dimensions in any documented way — likely a documentation gap.
2. **Overlap validity.** Each declared `axiom_overlap` value must reference a real dimension (`DIM-1` through `DIM-8`). Typos like `DIM-9` or `DIM-1A` are flagged.

**Severity rationale.** Pairing-contract violations don't break functionality; they erode the design-time interleaving experience over time. LOW severity reflects that. These findings never escalate the audit verdict beyond `NEEDS_ATTENTION` and never produce HIGH/MEDIUM unless other findings independently warrant them.

**Output format.** Findings carry `dimension: pairing-contract` (a meta dimension distinct from DIM-1..DIM-8) so reviewers can filter them when desired:

```json
{
  "severity": "LOW",
  "dimension": "pairing-contract",
  "skill": "exarchos:design-invariants",
  "description": "Paired skill declares pairs-with: axiom:design but no invariant has an axiom_overlap declaration.",
  "required_fix": "Add axiom_overlap: DIM-N to at least one invariant, or update the pairing-contract declaration to reflect that this skill is project-specific without dimension overlap."
}
```

## Error Handling

- **Empty scope (no files to assess):** Return "Nothing to assess — no files found in scope" with verdict CLEAN
- **Partial skill failure:** If one skill errors, continue with the remaining skills. Report the error in the output: "axiom:{skill} encountered an error: {message}. Results from other skills are still valid."
- **Scope validation:** Verify file/directory exists before starting. If not found: "Scope not found: {path}"

## Composition Guide

For details on how audit discovers skills, handles deduplication, and formats reports, see `@skills/audit/references/composition-guide.md`.

## References

- Dimension taxonomy: `@skills/backend-quality/references/dimensions.md`
- Finding format: `@skills/backend-quality/references/findings-format.md`
- Scoring model: `@skills/backend-quality/references/scoring-model.md`
- Composition details: `@skills/audit/references/composition-guide.md`
- Pairing contract: `@skills/backend-quality/references/pairing-contract.md`