audit

Files (1)

• SKILL.md

SKILL.md (first lines)

# Data Quality Audit

Run a 15-rule scan across all papers, organized into three severity levels: error (must-fix),
warning (should-fix), and info (nice-to-know). Produces a structured report showing which papers
have which issues, with fix guidance for each rule.

## Prerequisites

Papers must be in the library (`paper-ingest` skill). For rules checking concept/edge counts, the KG should be built (`kg-build` skill).

## Quick Start

```bash
drbrain audit
```

## What It Checks

**Error rules (2):** `missing_title`, `missing_md`
**Warning rules (8):** `missing_doi`, `missing_abstract`, `missing_year`, `missing_journal`,
  `missing_authors`, `short_md`, `empty_tree`, `low_concept_count`, `unresolved_env`
**Info rules (4):** `no_edges`, `placeholder_status`, `old_placeholder`, `duplicate_title`

## Common Patterns

**Quick health check before analysis:**
```bash
drbrain audit --severity error
```
If this returns issues, something is fundamentally broken — fix before running analysis.

**Full audit before `drbrain analyze` or `drbrain seed`:**
```bash
drbrain audit
```
Fix warnings first. Papers with missing abstracts, years, or journals produce noisy results.

**Find stale placeholders:**
```bash
drbrain audit --severity info --json | jq '.[] | select(.rule == "old_placeholder")'
```

**Workspace-specific audit:**
```bash
drbrain audit --workspace attention-methods
```

**Fix common issues:**
- `missing_md` / `empty_tree`: re-run `drbrain ingest`
- `missing_doi` / `missing_authors` / `missing_abstract`: run `drbrain repair --all`
- `low_concept_count` / `no_edges`: re-run `drbrain graph build`

## Examples

**Full audit with JSON for scripting:**
```bash
drbrain audit --severity info --json | jq 'group_by(.severity) | map({severity: .[0].severity, count: length})'
```

**Check a workspace for issues before submission:**
```bash
drbrain audit --workspace paper-draft --severity warning
```

## CLI Reference

| Command | What it does |
|---------|--------------|
| `drbrain audit` | Full audit at warning level (warnings + errors) |
| `drbrain audit --severity error` | Only errors |
| `drbrain audit --severity info` | All issues including info-level |
| `drbrain audit --json` | Machine-readable JSON output |
| `drbrain audit --workspace <name>` | Audit only papers in a workspace |
| `drbrain repair --all` | Auto-fix missing metadata from CrossRef/arXiv |

From UynajGI/drbrain, MIT.

More from UynajGI/drbrain