Claude Cursor Skill

verify-refs

Use when checking BibTeX reference records for missing fields, malformed identifiers, duplicate keys, or metadata mismatches before submission.

LLM Mart · 0 points · 3 views 0 listing impressions 0 install-command copies
Download yha9806-academic-writing-toolkit-.claude_skills_verify-refs-488622a.zip · 13 KB

Install

skills CLI npx skills add https://github.com/yha9806/academic-writing-toolkit/tree/main/.claude/skills/verify-refs
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install yha9806-academic-writing-toolkit@llmmart
Git git clone https://github.com/yha9806/academic-writing-toolkit.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole yha9806/academic-writing-toolkit collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

/verify-refs — Reference Authenticity Check

Running Python helpers

Choose the interpreter before running the examples. For a globally installed copy, use the private runtime recorded by its installer. In a checkout or linked workspace, use AWT_PYTHON when set, otherwise the toolkit's .venv (follow the skill directory link back to the toolkit): Scripts/python.exe on Windows, bin/python on macOS/Linux. Without that environment, check that python (Windows) or python3 (macOS/Linux) actually runs and has the helper's dependencies. Replace the example's python3 with that executable. In PowerShell, prefix a quoted executable with &; keep commands on one line and quote file paths.

Purpose

Check reference records for missing required fields, duplicate keys, malformed DOI values, malformed arXiv identifiers, and invalid URLs. The default mode is offline and deterministic.

Trigger Words

This skill activates on: verify refs, verify references, reference check, /verify-refs.

Workflow

  1. Identify the target Markdown or .bib file. If the user does not specify one, ask.
  2. Run: python3 .claude/skills/verify-refs/scripts/verify-refs.py --bib "{path}" --json
  3. Report issues by entry key and severity.
  4. For explicit metadata verification, run: python3 .claude/skills/verify-refs/scripts/verify-refs.py --bib "{path}" --json --online
  5. Use CrossRef for DOI metadata, Semantic Scholar as a secondary metadata source, and arXiv for preprint identifiers. Online checks must be explicit because they depend on network availability.
  6. In tests or offline review, use --metadata-dir "{dir}" to read CrossRef JSON, Semantic Scholar JSON, and arXiv Atom fixtures instead of live network calls.
  7. For a LaTeX manuscript, reconcile its citations with the .bib file both ways: python3 .claude/skills/verify-refs/scripts/reconcile-cites.py --bib "{bib}" --root "{repo}" --json "{main.tex}" ... It follows \input, \include and \subfile, reports keys cited but not defined and entries defined but cited nowhere, and treats \nocite{*} as citing everything. Macro definitions (\newcommand, \renewcommand, \providecommand, \def, \let ...) are skipped with their bodies, and a token starting with \ or # is never read as a key; a key written out in a definition body counts as cited. Exit 2 when nothing could be read.

Constraints

  1. Do not import project-specific reference rules from other repositories.
  2. Do not auto-fix reference records without user approval.
  3. Keep output plain Markdown with no emoji.
  4. Treat CrossRef, Semantic Scholar, and arXiv as verification sources, not as citation style authorities.
Files (academic-writing-toolkit)
  • scripts
    • reconcile-cites.py 18.2 KB
      Loading…
    • verify-refs.py 17.3 KB
      Loading…
  • SKILL.md 2.9 KB
    ---
    name: verify-refs
    description: Use when checking BibTeX reference records for missing fields, malformed identifiers, duplicate keys, or metadata mismatches before submission.
    allowed-tools: Read, Glob, Bash, WebFetch
    ---
    
    # /verify-refs — Reference Authenticity Check
    
    ## Running Python helpers
    
    Choose the interpreter before running the examples. For a globally installed
    copy, use the private runtime recorded by its installer. In a checkout or linked
    workspace, use `AWT_PYTHON` when set, otherwise the toolkit's `.venv` (follow the
    skill directory link back to the toolkit): `Scripts/python.exe` on Windows,
    `bin/python` on macOS/Linux. Without that environment, check that `python`
    (Windows) or `python3` (macOS/Linux) actually runs and has the helper's dependencies.
    Replace the example's `python3` with that executable. In PowerShell, prefix a
    quoted executable with `&`; keep commands on one line and quote file paths.
    
    ## Purpose
    
    Check reference records for missing required fields, duplicate keys, malformed DOI values, malformed arXiv identifiers, and invalid URLs. The default mode is offline and deterministic.
    
    ## Trigger Words
    
    This skill activates on: `verify refs`, `verify references`, `reference check`, `/verify-refs`.
    
    ## Workflow
    
    1. Identify the target Markdown or `.bib` file. If the user does not specify one, ask.
    2. Run:
       `python3 .claude/skills/verify-refs/scripts/verify-refs.py --bib "{path}" --json`
    3. Report issues by entry key and severity.
    4. For explicit metadata verification, run:
       `python3 .claude/skills/verify-refs/scripts/verify-refs.py --bib "{path}" --json --online`
    5. Use CrossRef for DOI metadata, Semantic Scholar as a secondary metadata source, and arXiv for preprint identifiers. Online checks must be explicit because they depend on network availability.
    6. In tests or offline review, use `--metadata-dir "{dir}"` to read CrossRef JSON, Semantic Scholar JSON, and arXiv Atom fixtures instead of live network calls.
    7. For a LaTeX manuscript, reconcile its citations with the `.bib` file both ways:
       `python3 .claude/skills/verify-refs/scripts/reconcile-cites.py --bib "{bib}" --root "{repo}" --json "{main.tex}" ...`
       It follows `\input`, `\include` and `\subfile`, reports keys cited but not defined and entries defined but cited
       nowhere, and treats `\nocite{*}` as citing everything. Macro definitions (`\newcommand`, `\renewcommand`,
       `\providecommand`, `\def`, `\let` ...) are skipped with their bodies, and a token starting with `\` or `#` is never
       read as a key; a key written out in a definition body counts as cited. Exit 2 when nothing could be read.
    
    ## Constraints
    
    1. Do not import project-specific reference rules from other repositories.
    2. Do not auto-fix reference records without user approval.
    3. Keep output plain Markdown with no emoji.
    4. Treat CrossRef, Semantic Scholar, and arXiv as verification sources, not as citation style authorities.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.