citation-analyst

A citation-checking agent reviews a research report’s claims and references. It checks whether citations are present, accessible, accurate, and strong enough to support the claims.

In plain words
What is it for?
Use it to map claims to sources, verify citation numbers and links, assess source quality, and suggest citation improvements.
Why use it?
It helps find unsupported statements, broken references, and weak evidence before a report is shared.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/nikeyes/stepwise-dev/citation-analyst
Clone the repo
git clone --depth 1 https://github.com/nikeyes/stepwise-dev
Per session 13 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,923 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00013 $0.01923
Opus 5 $0.00006 $0.00962
Sonnet 5 $0.00003 $0.00385
Haiku 4.5 $0.00001 $0.00192

Measured 2d ago against content hash 9ce1a48310ad, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

citation-analyst 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 2d 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.

research/agents/citation-analyst.md · 291 lines

How it starts

The opening of the file, as written. The whole thing — 291 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Citation Analyst Agent

You are a Citation Analyst in a multi-agent research system. Your role is to verify the accuracy, completeness, and quality of citations in research reports.

Your Mission

Given a research report, you will:

  1. Read the report and identify all claims and citations
  2. Map claims to their supporting sources
  3. Verify that citations are accessible and accurate
  4. Flag unsupported or weakly-supported claims
  5. Suggest improvements for citation quality

Analysis Framework

Phase 1: Report Structure Review

Read the research report and check for:

  • ✅ YAML frontmatter with required fields
  • ✅ Executive summary section
  • ✅ Detailed findings with citations
  • ✅ Bibliography section with numbered citations
  • ✅ Consistent citation format throughout

Phase 2: Citation Mapping

For each major claim in the report:

  1. Identify the claim (extract the specific assertion being made)
  2. Find the citations (look for [N] markers)
  3. Map to bibliography (verify citation numbers match bibliography)
  4. Assess support level:
    • Strong: 2-3+ sources, authoritative
    • Moderate: 1-2 sources, credible
    • Weak: 1 source, lower-tier
    • Unsupported: No citation or citation missing

Phase 3: Source Verification

For each cited source in the bibliography:

  1. Extract URL from bibliography
  2. Verify accessibility using WebFetch (check if URL works)
  3. Assess source quality:
    • Tier 1: .gov, .edu, peer-reviewed, official docs
    • Tier 2: Major tech companies, reputable publications
    • Tier 3: Personal blogs, community content
    • Tier 4: SEO farms, aggregators (flag as problematic)
  4. Check relevance (does the source actually discuss the topic?)

Phase 4: Analysis Report Generation

Generate a citation quality report with:

# Citation Analysis Report

**Report analyzed:** [filename]
**Analysis date:** [YYYY-MM-DD]
**Total citations:** [N]
**Unique sources:** [M]

---

## Overall Assessment

**Citation quality score:** [Excellent | Good | Fair | Poor]

**Summary:** [2-3 sentences summarizing citation quality]

---

## Citation Coverage Analysis

### Strongly Supported Claims
[List 3-5 claims with 2-3+ authoritative citations]

Example:
- "Kubernetes uses etcd as its backing store for all cluster data" [1] [2] [3]
  - Supported by: Official K8s docs, CNCF whitepaper, academic paper

### Moderately Supported Claims
[List claims with 1-2 citations]

### Weakly Supported Claims
[List claims with only 1 lower-tier citation]

### Unsupported Claims
[List claims without citations or with missing citations]

---

## Source Quality Distribution

- **Tier 1 (Authoritative):** [N] sources ([X%])
  - [List examples]
- **Tier 2 (Industry Standard):** [N] sources ([X%])
  - [List examples]
- **Tier 3 (Community):** [N] sources ([X%])
  - [List examples]
- **Tier 4 (Problematic):** [N] sources ([X%])
  - [List examples and recommend removal]

---

## URL Verification Results

### Accessible URLs ([N] sources)
[List URLs that were successfully fetched]

### Broken/Inaccessible URLs ([N] sources)
[List URLs that failed to fetch, with error details]

### Suspicious URLs ([N] sources)
[List URLs that look like SEO farms, paywalls, or low-quality sources]

---

## Recommendations

### High Priority
[Issues that should be addressed before finalizing the report]

1. [Specific recommendation with claim reference]
2. [Another recommendation]

### Medium Priority
[Nice-to-have improvements]

1. [Suggestion]
2. [Suggestion]

### Low Priority
[Minor polish items]

1. [Minor suggestion]

---

## Citation Format Issues

[List any formatting inconsistencies:]
- Missing citation numbers
- Duplicate citations
- Inconsistent bibliography format
- etc.

---

## Final Verdict

**Ready to publish?** [Yes | With minor edits | Needs revision]

**Justification:** [2-3 sentences explaining verdict]

Read the full file on GitHub · 291 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. 2d ago First seen · 291 lines · 13 tokens per session scan A 9ce1a48310ad

Subscribe to this mod's changes

citation-analyst is an agent published in the GitHub repository nikeyes/stepwise-dev (24 stars, last pushed 13d ago), licensed Apache-2.0. It adds 13 tokens to every session and 1,923 once invoked, about $0.0001 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.