Paper Quality Citation Intelligence

When not to use

Do not use this skill when:

• the user only wants a plain summary of the paper
• the user wants a full expert review or peer-review-style critique
• the task is a full systematic literature review
• the user wants detailed theorem verification or deep technical proof checking
• there is no identifiable paper or paper metadata to work from

If the user only wants a simple summary, provide a normal summary instead of this triage workflow.

Required Inputs

Prefer at least one of the following:

• paper title
• DOI
• paper URL
• abstract
• venue and year

Optional but helpful:

• the user’s research topic or goal
• the user’s concern about the paper
• notes about what the user wants to evaluate, such as novelty, rigor, or importance

If the user provides only partial information, proceed cautiously and lower confidence where needed.

Output Requirements

The output must be:

• structured
• concise but informative
• easy to scan in 1–2 minutes
• actionable for a research student

Workflow

Follow this workflow in order.

Step 1. Identify the paper

Determine the best available paper identity from the user’s input or the tool output.

If the identity is uncertain:

• mention the ambiguity
• avoid strong claims
• continue only with appropriate caution

Step 2. Establish the evaluation context

Infer the likely topic and the user’s likely research purpose if provided.

Examples:

• core technical anchor
• background understanding
• implementation inspiration
• related-work coverage

This context affects whether a paper is worth deep reading.

Step 3. Assess publication signal

Evaluate:

• publication type: journal, conference, workshop, preprint, or unknown
• venue visibility or likely centrality in the field, if known
• whether the paper is recent enough that low citations are expected

Do not overstate venue quality if uncertain.

Step 4. Interpret citation signal beyond raw count

Use the available evidence to judge whether the paper appears to be:

• mostly cited as background
• reused methodologically
• extended by later work
• used as a benchmark or comparison target
• disputed or contrasted
• too weakly evidenced to judge

Use references/citation-signals.md for interpretation rules.

Step 5. Assess paper quality using the rubric

Use the rubric in references/evaluation-rubric.md to evaluate:

• relevance
• contribution clarity
• methodological strength
• evaluation rigor
• likely weaknesses
• follow-up value

Step 6. Rescue value from references if needed

If the paper seems weak, shallow, or uncertain, do not stop there.

Inspect its references or related papers and identify:

• foundational papers
• stronger related works
• more recent alternatives
• contrasting or competing works

Step 7. Produce a structured report

Return the final report using the template in references/output-template.md.

Always end with exactly one final recommendation:

• Read deeply
• Skim
• Background only
• Skip, but inspect references

Confidence and Uncertainty Policy

Lower confidence when:

• the paper identity is uncertain
• venue metadata is missing
• citation analytics are incomplete
• the paper is very recent
• only an abstract is available
• the user’s research goal is unclear

When confidence is low:

• say what is missing
• avoid absolute judgments
• explain why the recommendation is tentative

References

• references/citation-signals.md
• references/evaluation-rubric.md
• references/output-template.md

Examples

See:

• examples/example-input-1.md
• examples/example-output-1.md