# Ac Systematic Review Conduct

> Running a systematic review as a protocol before it is a document: the question in PICO form, eligibility fixed before the search, the search log and coverage grid, dual screening and inter-rater agreement, extraction forms and piloting, risk-of-bias assessment with a named tool, grey literature and trial registers, the PRISMA flow kept live, protocol registration and amendments, and the conduct audit trail. It runs the review's machinery; it never writes the report and never decides the conclusion. Use when asking "systematic review protocol", "PICO question", "dual screening", "kappa", "extraction form", "risk of bias tool", "PROSPERO", "search strategy log", "grey literature", "PRISMA flow diagram", "scoping versus systematic", "Cochrane handbook". Not for writing the review as a document (ac-academic-writing), not for open-corpus research hunts without a protocol (ac-deep-researcher), not for attacking a finished corpus (ac-adversarial-corpus-challenge).

- Skill: `jiayaoqijia/ac-systematic-review-conduct` (Agent Skill, multi-file: 13 files)
- Install (CLI): `npx get-skill add jiayaoqijia/ac-systematic-review-conduct`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jiayaoqijia/ac-systematic-review-conduct/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- License: MIT
- Author: jiayaoqijia (https://skillmd.com/u/jiayaoqijia)
- Updated: 2026-11-16
- Page: https://skillmd.com/skills/jiayaoqijia/ac-systematic-review-conduct

---


# Systematic review conduct

## Overview

A systematic review differs from a literature summary in one structural way: the method is written before the evidence is seen, and the review's credibility rests on whether that protocol survived contact with the results. The question is framed in advance, the criteria that decide inclusion are fixed before the search runs, and every departure from the plan is an amendment with a reason — not a quiet adjustment. What gets written up afterwards is the residue of what was conducted.

This skill covers the conduct: framing the question, specifying the search and proving its coverage, screening in duplicate, extracting with a piloted form, assessing bias with a named tool rather than a feeling, and keeping the records that make the whole thing auditable. It is deliberately narrow — the writing of the review is a separate skill, and so is the open-corpus hunt of a question that never claimed to be systematic.

Primary sources: <https://training.cochrane.org/handbook> and <https://equator-network.org/>.

## When to use

- When running a systematic or scoping review under a declared protocol.
- When a review's search, screening, or bias assessment must be reproducible.
- When registering a protocol or amending one mid-stream.
- When the review's inclusion decisions will be contested.

## When NOT to use

- For writing the review as a document (use ac-academic-writing).
- For open-ended research hunting with no protocol (use ac-deep-researcher).
- For adversarially attacking an existing corpus (use ac-adversarial-corpus-challenge).

## Subskill Index

| Leaf | Domain segment | Scope (one line) |
|---|---|---|
| `subskills/the-question-and-eligibility.md` | `research-operations/question-and-eligibility` | PICO framing and eligibility criteria written before the search. |
| `subskills/search-and-coverage.md` | `research-operations/search-and-coverage` | Database selection, the search log, and the coverage grid that defends it. |
| `subskills/screening-and-agreement.md` | `research-operations/screening-and-agreement` | Dual screening, disagreement resolution, and inter-rater statistics. |
| `subskills/extraction-forms.md` | `research-operations/extraction-forms` | Designing, piloting, and versioning the data extraction form. |
| `subskills/risk-of-bias.md` | `research-operations/risk-of-bias` | Named bias tools, per-domain judgments, and the assessment record. |
| `subskills/grey-literature-and-registers.md` | `research-operations/grey-literature-and-registers` | What the databases miss, and the registers that catch it. |
| `subskills/the-prisma-record.md` | `research-operations/prisma-record` | The flow diagram as a counting record, kept while the review runs. |
| `subskills/protocol-registration.md` | `research-operations/protocol-registration` | Registering the protocol before results are known, and amending honestly. |
| `subskills/conduct-record.md` | `research-operations/conduct-record` | The audit trail: protocol version, amendments, disagreements, tool changes. |

## Status and honest limits

Conduct is what makes a review systematic rather than authoritative-sounding; if the protocol changed after evidence was seen, the record says so. This skill never writes the conclusions and never judges a study's science — risk of bias is about internal validity of the evidence process, not the truth of the finding. A review conducted perfectly can still be wrong if the underlying literature is thin; the record says what was reviewed, never that enough evidence existed.

