grounded-citations

grounded-citations is a skill for Claude Code from mronkko/claude-academic-research. It costs 110 tokens per session (1,834 once invoked), scanned A, original, MIT.

A procedure for adding a citation to academic writing or explaining what a research paper says. It consults an external source and uses it only when that source supports the claim.

In plain words
What is it for?
Finding support for a statement, adding citations to academic prose, attributing findings, and summarizing a paper from its source.
Why use it?
It reduces unsupported statements and citations based only on memory. If no source supports a claim, the procedure can leave the claim out.

Skill for Claude Code

Written for Claude Code: ${CLAUDE_PLUGIN_ROOT variable.

Runs only inside its plugin — its command needs a path that Claude Code sets for a plugin’s own hooks and for nothing else. Install the plugin, not this.

Part of the academic-research plugin — 11 skills, 1 plugin shipped together

Good fit Finding support for a statement, adding citations to academic prose, attributing findings, and summarizing a paper from its source.

Compare 6 skills from other repositories ↓
Install

Getting it into your agent

This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.

Claude Code
/plugin marketplace add mronkko/claude-academic-research
Claude Code
/plugin install academic-research

Made for: Claude Code.

Or install academic-research, the plugin that ships this one along with the rest of its 11 skills, 1 plugin.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for grounded-citations

README.md
[![agentmods](https://agentmods.dev/badge/skills/mronkko/claude-academic-research/grounded-citations/github.svg)](https://agentmods.dev/skills/mronkko/claude-academic-research/grounded-citations)
Your own site
<a href="https://agentmods.dev/skills/mronkko/claude-academic-research/grounded-citations"><img src="https://agentmods.dev/badge/skills/mronkko/claude-academic-research/grounded-citations/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for grounded-citations

Your own site · 80×15
<a href="https://agentmods.dev/skills/mronkko/claude-academic-research/grounded-citations"><img src="https://agentmods.dev/badge/skills/mronkko/claude-academic-research/grounded-citations.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 110 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,834 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00110 $0.01834
Opus 5 $0.00055 $0.00917
Sonnet 5 $0.00022 $0.00367
Haiku 4.5 $0.00011 $0.00183

Measured 11d ago against content hash e582795e4e42, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

grounded-citations scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/grounded-citations/SKILL.md · 145 lines

How it starts

The opening of the file, as written. The whole thing — 145 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Grounded citations

Glossary: unfamiliar with BBT, MCP, DOI? See skills/_glossary.md for one-line definitions of every acronym this skill uses.

Insertion vs verification. This skill governs inserting a citation during drafting — it can drop a claim if no source supports it. The audit-time counterpart is verifying-citations (loaded by fact-check and by critic-loop's evidence critic), which decides whether an existing @citekey in a manuscript is honestly supported by its source. Both demand externalised consultation; their escalation triggers differ: grounded-citations escalates on recency / context staling (a Zotero child note beats a faded context-window memory); verifying-citations escalates on claim type (a quoted passage or specific β always requires the full text, regardless of how the abstract reads). Same goal, different decision axis.

Bootstrap (first run in this project)

Before applying the rules below, check that this skill's regression tests are installed in the project:

python3 "${CLAUDE_PLUGIN_ROOT:-.}/scripts/setup/check_project_scaffold.py" \
    scripts/test_common.py scripts/test_citations.py

If the output lists missing files, install them:

python3 "${CLAUDE_PLUGIN_ROOT:-.}/scripts/setup/install_templates.py" \
    test_common.py:scripts/test_common.py \
    test_citations.py:scripts/test_citations.py

Then tell the user what was installed and flag that the top of scripts/test_citations.py has project-specific paths (manuscript, references.bib, optional coded_papers.csv) they should review.

Core rule

Every citation the agent inserts into academic prose must satisfy all four of these requirements. They are conjunctive — failing any one means the citation may not be made.

  1. In Zotero. The paper is in the project's Zotero library. If not, add it via mcp__zotero__zotero_add_item (source_type="doi", or "url" when no DOI exists) before the citation is written.
  2. BBT key. The [@citekey] in prose is the Better BibTeX key fetched from Zotero via mcp__zotero__zotero_get_item_metadata with format="bibtex". Never hand-craft keys (Smith2019-style); never fabricate; never write to Zotero's Extra field to pin a key.
  3. Externalised consultation. At citation time, the paper's content is available either as a fresh MCP response (abstract, notes, or full-text retrieved in the current turn or an adjacent recent turn) or as a Zotero child note read via mcp__zotero__zotero_get_notes. Context-window recall alone is not sufficient: remembering an abstract read 500K tokens ago is not grounding — either re-fetch it or read the note. Operational rule: if the abstract or fulltext was read more than ~5 conversational turns ago, or if context has been compacted in between, re-fetch via MCP or read the stored Zotero note. The boundary is "could the same words still be in your working context"; when in doubt, re-fetch.
  4. Claim support. The consulted content visibly supports the attributed claim. If nothing you have consulted supports the claim, drop the claim. Do not paper over; do not flag for later; do not keep a speculative citation. Remove it from prose, or replace the attribution with a source that does support the claim.

Read the full file on GitHub · 145 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 11d ago First seen · 145 lines · 110 tokens per session scan A e582795e4e42

Subscribe to this mod's changes

grounded-citations is a skill published in the GitHub repository mronkko/claude-academic-research (23 stars, last pushed 6d ago), licensed MIT. It adds 110 tokens to every session and 1,834 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

local-ai-agents

Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…

microsoft/ai-agents-for-beginners · 200 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

insight-error-page

Write or audit an insight-kind error page for the Next.js dev overlay. Use when creating a new errors/ .mdx page, auditing an existing one, or checking that a page matches the framework fix cards. Covers page structure, title alignment, FixCard cards with Copy prompt button, code snippets, terminology verification…

vercel/next.js · 83 tokens

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens

next-partial-prefetching-adoption

Turn on Partial Prefetching in a Next.js app and work through the insights it surfaces. Use when the user wants to enable or adopt Partial Prefetching, flip the partialPrefetching flag, opt routes in with export const prefetch = 'partial', audit Link prefetch={true} behavior, preserve existing prefetched UI with…

vercel/next.js · 103 tokens