# Quality Assurance

> Verification strategy that independently re-checks a paper bundle's precision (no unsupported claims) and recall (nothing important omitted) against the original source, before any article drafting begins, then re-checks the finished draft for drift. Use this after deep-read produces a bundle and before audience-first-writing drafts an article — this is the mandatory middle strategy of the paper-reading pipeline.

- Skill: `yogsoth-ai/quality-assurance` (Agent Skill)
- Install (CLI): `npx skillmds@latest add yogsoth-ai/quality-assurance`
- Raw SKILL.md: https://api.skillmd.com/api/skills/yogsoth-ai/quality-assurance/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: DevOps & Infra
- Author: yogsoth-ai (https://skillmd.com/u/yogsoth-ai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/yogsoth-ai/quality-assurance

---


# Quality Assurance

**Purpose**: Catch commission errors (wrong claims) and omission errors (missing content) in the bundle before writing, and drift errors (rewrite diverged from bundle) after writing — via independent re-querying of the source, never by checking the bundle against itself.

**When to use**: Always, between deep-read and audience-first-writing — the second strategy in the fixed pipeline sequence (see `paper-reading/ENTRY.md`).

## Available Tactics

- `dual-gate-verification` — the only tactic in v1 (exhaustive strict path). `lightweight-spot-check` (sampled) and `human-in-loop` (high-stakes) are deferred to v2.

## Execution Guidance

Invoke `dual-gate-verification` directly. There is no tactic-selection decision in v1 since only one tactic exists.

## Failure Routing

See `failure-routing` sop — this strategy never proceeds past a detected precision or recall failure; it always routes back to `deep-read` first.