related-work-scout

name	related-work-scout
description	Use when you want Cursor Agent to collect structured candidate related-work paper cards into assets/notes/agent_artifacts/literature/provider_output.json for your own bibliography and drafting workflow.

Related work scout

This skill is for constrained literature scouting, not prose writing.

Use it when:

• the goal is to collect candidate related-work papers
• you will curate references.bib and prose yourself
• the model should behave like a researcher or curator, not an author

Do not use it to:

• draft the Related Work section in one shot without human review
• approve papers automatically
• claim novelty
• write anything outside the artifact file

Output contract

Write exactly one JSON artifact to:

assets/notes/agent_artifacts/literature/provider_output.json

Read the schema in references/provider-output-schema.json before producing output.

Allowed top-level shapes:

{"papers": []}

or a bare JSON array.

Each paper must stay within evidence you can justify from the available project context and the papers you inspected. If a field is unknown, use an empty string or empty list. Do not invent DOI, venue, year, or claims.

Workflow

1. Read assets/notes/project.md and any assets/refs/*.json if present.
2. Read references/provider-output-schema.json.
3. Produce candidate paper cards only.
4. Save them to assets/notes/agent_artifacts/literature/provider_output.json.
5. Stop.

Constraints

• Prefer 8 to 20 papers unless the available evidence is thinner.
• Separate same-setting work from adjacent-setting work in relation_to_us.
• Set status to candidate.
• Set usable_in to ["related_work"].
• Keep method_tags short and concrete.
• Use stable paper_id values such as P-... or other unique ids.
• Do not include Markdown, explanation text, or comments in the artifact.

Prompts

For Codex, reuse references/codex-prompt.md.

For Claude, reuse references/claude-prompt.md.