test-audit-pipeline

name	test-audit-pipeline
version	0.8.0
description	One-command test health check that orchestrates coverage collection, TDD quality scoring, and redundancy triage into a unified pipeline. Runs independent stages in parallel for ~40% faster execution. Produces a single merged report with rubric scores, triage decisions, and actionable next steps.

Test Audit Pipeline

Overview

Orchestrates test-quality-assurance and test-redundancy-triage into one workflow with shared coverage collection and a unified report.

Prerequisites

• Sibling skills installed: test-quality-assurance (audit_test_quality.py) and test-redundancy-triage (triage_redundancy.py).
• Python 3.10+; pytest; coverage is optional for coverage integration.

Quick Start

python ~/.agents/skills/test-audit-pipeline/scripts/audit_pipeline.py \
  --root /path/to/repo \
  --python .venv/bin/python \
  --suite tests/test_api.py \
  --suite tests/test_core.py \
  --source-prefix src/mypackage/ \
  --internal-import-pattern "from\s+mypackage\.(core|impl)\s+import" \
  --public-hint "compute(" \
  --out-dir /tmp/full_audit \
  --env NUMBA_DISABLE_JIT=1

Parallelism

The pipeline stages have the following dependency graph:

Stage 1: Collect coverage (pytest --cov → coverage.json)
    ↓
Stage 2a: TQA audit (uses coverage.json)     ← PARALLEL
Stage 2b: Redundancy triage                  ← PARALLEL
    ↓
Stage 3: Unified report (merges TQA + triage results)

Stage 2a and 2b run concurrently with ThreadPoolExecutor; coverage must finish before TQA, and the report waits for both. Agent orchestrators may delegate TQA and triage to separate subagents.

Common Runs

Full audit:

python scripts/audit_pipeline.py \
  --root /path/to/repo \
  --python .venv/bin/python \
  --suite tests/test_api.py \
  --comparator-suite tests/test_integration.py \
  --source-prefix src/pkg/ \
  --out-dir /tmp/audit \
  --env NUMBA_DISABLE_JIT=1

Baseline comparison:

python scripts/audit_pipeline.py \
  --root /path/to/repo \
  --python .venv/bin/python \
  --suite tests/ \
  --out-dir /tmp/audit \
  --tqa-baseline /tmp/previous_audit/tqa_report.json \
  --env NUMBA_DISABLE_JIT=1

Coverage-only mode:

python scripts/audit_pipeline.py \
  --root /path/to/repo \
  --python .venv/bin/python \
  --out-dir /tmp/audit \
  --skip-triage \
  --env NUMBA_DISABLE_JIT=1

Outputs

out-dir/
├── pipeline_report.md          # Unified human-readable report
├── pipeline_summary.json       # Machine-readable merged results
├── coverage.json               # Raw coverage data (when collected)
├── tqa_report.json             # TQA JSON output
├── tqa_report.md               # TQA markdown output
└── triage/                     # Triage sub-directory
    ├── candidate_validation.md
    ├── candidate_validation.csv
    ├── confidence_gate_matrix.csv
    └── ... (all triage outputs)

Framework Notes

• Numba: Pass --env NUMBA_DISABLE_JIT=1 to get truthful coverage
• Django: Pass --env DJANGO_SETTINGS_MODULE=myproject.settings
• General: Any env var needed by the test suite should be passed via --env

Limits

• Sibling skill discovery is relative to the test-quality-assurance and test-redundancy-triage install directories; use --tqa-script and --triage-script for non-standard layouts.
• Coverage collection runs all matched tests; use --test-marker for filtering.
• The unified report is additive; individual reports are still written.