# Dr Cook:citation Checker

> Validate and correct citation formatting in academic manuscripts. Use when checking references, fixing citation style errors, validating in-text citation matching, auditing self-citation rates, or formatting to Vancouver, APA, or GB/T 7714-2015. Triggers on: check citations, fix references, validate reference list, citation errors, self-citation audit, Vancouver/APA/GB/T 7714, 检查参考文献, 格式化引用, 自引检查. Do NOT trigger on: literature review, paper download, manuscript writing, English polishing, or peer review comments.

- Skill: `wen-chen/dr-cook-citation-checker` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add wen-chen/dr-cook-citation-checker`
- Raw SKILL.md: https://api.skillmd.com/api/skills/wen-chen/dr-cook-citation-checker/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: wen-chen (https://skillmd.com/u/wen-chen)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/wen-chen/dr-cook-citation-checker

---


# citation-checker

## 1. Overview

Validates reference lists and in-text citations against Vancouver, APA 7th, and GB/T 7714-2015. Identifies formatting errors, orphaned citations, and self-citation rate violations. Operates standalone (user pastes text) or in pipeline mode (receives `raw_text` and `references` from upstream). Output: numbered issues list with severity tags and a corrected reference list. Bilingual input supported. Domains: tcm, bioinformatics, clinical, pharmacology.

## 2. Parameters

### Required

| Parameter | Values | Description |
|---|---|---|
| `citation_style` | `vancouver` \| `apa` \| `gb-t-7714` | The target citation standard. If inherited from upstream `context_output`, skip asking. |
| `text` | string | The manuscript body, reference list, or both. May be pasted directly or inherited from `context_output.raw_text`. |

### Optional

| Parameter | Values | Description |
|---|---|---|
| `check_self_citation` | bool (default: `false`) | Whether to audit the reference list for self-citations by the manuscript's authors. |
| `author_name` | string | Full name or last name + initials of the submitting author. Required when `check_self_citation = true`. |
| `domain` | `tcm` \| `bioinformatics` \| `clinical` \| `pharmacology` | Research domain. Used for context only; does not alter citation rules. Inherit from upstream if available. |

### Parameter collection rule

Collect one missing required parameter at a time. If `citation_style` and `text` are both available from upstream `context_output`, proceed without prompting. If `check_self_citation = true` but `author_name` is missing, ask for it before Step 6.

## 3. Workflow

**Step 1 — Check for upstream context_output.**

Inspect active `context_output`. Inherit `parameters.citation_style` → `citation_style`, `parameters.domain` → `domain`, `raw_text` → `text`, `references` → structured reference list. Record inherited fields for the output header. If no upstream context, proceed to Step 2.

**Step 2 — Collect missing parameters.**

- If `citation_style` not inherited, ask: "Which citation style: Vancouver, APA 7th, or GB/T 7714-2015?"
- Resolve `text` using the first available source: (1) `context_output.raw_text` if present; (2) text pasted directly in the user's message; (3) content the user references as "the file I shared" or a previously-displayed block — acknowledge and use it. If none apply, ask: "Please paste the reference list and/or manuscript text you'd like me to check."
- Collect one parameter at a time.
- If `check_self_citation = true`, ask for `author_name` immediately after confirming.
- If the user has not mentioned self-citation checking, briefly offer it after confirming `citation_style` and `text` are in hand: "Would you also like me to check the self-citation rate? If so, please provide your name as it appears in your publications."

**Step 3 — Load citation format rules.**

Load `references/citation-formats.md`. Extract the rule set for the selected `citation_style` (field-order, author format, DOI format, in-text regex patterns). Keep active through Steps 4 and 5.

**Step 4 — Run format consistency check.**

Parse the reference list from `text` (or `references` array if inherited). For each entry apply the rule set from Step 3: field order, author format and "et al" threshold, journal name (abbreviated vs. full), DOI format, punctuation, and GB/T 7714 document type identifier.

Severity: ERROR (missing required field, wrong field order, wrong type identifier); WARNING (wrong punctuation, DOI format mismatch, journal name not abbreviated when required); INFO (stylistic inconsistency not violating the standard).

**Step 5 — Run in-text citation matching.**

Extract all in-text markers using patterns from `references/citation-formats.md`. Style-specific rules apply:

- **Vancouver and GB/T 7714:** extract numeric markers using the bracket-number regex patterns from citation-formats.md.
- **APA:** build a surname-year lookup key for each reference (e.g., "Smith 2022"). Scan manuscript text for `(Author, Year)` patterns. When "et al." appears, match on first-author surname only (applies when ≥3 authors). Flag references with no match in text (orphan). Flag in-text citations with no reference list entry (dangling).

Compare in both directions:
- Cited in-text but absent from list → ERROR: "In-text citation [X] has no corresponding reference list entry."
- In list but never cited in-text → WARNING: "Reference [X] is never cited in the text."

**Step 6 — Self-citation audit (conditional).**

Run only if `check_self_citation = true`. Load `references/self-citation-norms.md`. Match `author_name` (surname + initials, fuzzy, case-insensitive) against each reference's author fields. Calculate rate: `self_citations / total_references × 100%`. Apply thresholds from norms file (defaults: > 30% ALERT; > 20% WARNING; ≤ 20% INFO). List matched references by number.

**Step 7 — Output.**

Present output per Section 4. Write `context_output` per Section 5.

## 4. Output Format

Output consists of two mandatory sections, presented in this order.

**Section A — Issues Found**

```
## Issues Found

[N issue(s) detected]

1. [SEVERITY] Reference N: description of issue.
```

If no issues are found: state "No citation issues found." then proceed to Section B.

**Section B — Corrected Reference List**

Echo the full reference list with all corrections applied inline. Number each entry. If a correction was applied, append `[corrected]` in italics after the entry on the same line for traceability.

If `check_self_citation` was run, append a **Self-Citation Summary** block after Section B showing: total references, self-citation count, rate percentage, severity level, and list of self-cited reference numbers.

## 5. context_output

### Reads from upstream

| Field | Source module | Usage |
|---|---|---|
| `parameters.citation_style` | paper-writer, paper-reviewer | Skip asking user for style |
| `parameters.domain` | any upstream module | Contextual information |
| `parameters.target_journal` | paper-writer, cover-letter-writer | Informational; does not alter rules |
| `raw_text` | paper-writer, paper-reviewer, rebuttal-writer | Used as `text` input |
| `references` | literature-reviewer, literature-downloader | Used as structured reference list |

### Writes to output

```json
{
  "module": "citation-checker",
  "summary": "string — one sentence: references checked, error/warning counts, self-citation rate if applicable.",
  "key_findings": [
    "string — one entry per ERROR-severity issue found"
  ],
  "references": [
    {
      "pmid": "string | null",
      "doi": "string | null",
      "title": "string",
      "authors": "string",
      "year": "integer"
    }
  ],
  "parameters": {
    "citation_style": "vancouver | apa | gb-t-7714",
    "domain": "tcm | bioinformatics | clinical | pharmacology | null",
    "self_citation_rate": "float | null — percentage value, e.g. 25.0; null if check_self_citation was false"
  },
  "status": "success | partial | failed",
  "error_message": "string | null — null on success; error description on failure"
}
```

- `key_findings`: ERROR-severity issues only; WARNING and INFO are in output text only.
- `references`: corrected reference list; replaces any upstream `references` array.
- `status`: `"success"` — completed; `"partial"` — some references unparseable; `"failed"` — no valid reference list found.

### Downstream readers

`references` (corrected) → cover-letter-writer, grant-writer. `summary` and `key_findings` → paper-reviewer (if looping back for re-check).

## 6. References

- `citation-formats.md` — field-order rules, author format, DOI format, in-text regex patterns, and error checklists per style.
- `self-citation-norms.md` — rate thresholds by article type and journal tier, identification algorithm, and mitigation strategies.