Writer agent

Produces structured paper-style drafts from research notes and synthesis files.

The Writer is Feynman’s writing subagent. It takes research files produced by the Researcher and turns them into clear, structured documents — briefs, literature reviews, paper drafts, comparison reports, and summaries. The Writer is always the last substantive agent to run before the Verifier post-processes the draft.

Role

The Writer synthesizes a collection of extracted claims and citations into coherent prose. It understands academic conventions: claims are attributed to sources, methodology sections are described with sufficient detail for reproduction, results are presented with appropriate qualifiers, and limitations are discussed honestly.

The Writer works only from the evidence in the supplied research files. It does not introduce claims, tools, or sources that are not present in those inputs.

Input

The Writer receives:

• Research files produced by the Researcher (e.g., <slug>-research-web.md, <slug>-research-papers.md)
• Synthesis notes or comparison matrices prepared by the lead agent
• A specified output path and document type

Output

The Writer saves its main artifact to the path specified by the lead agent. Default paths:

Document type	Output path
Research brief	outputs/.drafts/<slug>-draft.md
Paper-style draft	papers/<slug>.md or outputs/.drafts/<slug>-draft.md
Final output	outputs/<slug>.md or papers/<slug>.md

Document types

The Writer handles several document formats:

• Research briefs — Concise summaries of a topic with key findings, produced by the deep research workflow
• Literature reviews — Survey-style documents mapping consensus, disagreement, and open questions across a field
• Paper drafts — Full academic papers with abstract, introduction, body sections, discussion, and open questions
• Comparison reports — Structured analyses of how multiple sources agree and differ
• Summaries — Condensed versions of longer documents or multi-source findings

Output structure

Every draft follows this structure:

# Title

## Executive Summary
2-3 paragraph overview of key findings.

## Section 1: ...
Detailed findings organized by theme or question.

## Section N: ...
...

## Open Questions
Unresolved issues, disagreements between sources, gaps in evidence.

Visuals

When research contains quantitative data, the Writer uses the pi-charts package to embed charts in the draft. For architectures, pipelines, or multi-step processes, it uses Mermaid diagrams. Visuals are only added when they materially improve understanding — not for decoration. Every visual has a descriptive caption and references the data it is based on.

Integrity rules

The Writer follows strict constraints to preserve evidence integrity:

1. Write only from supplied evidence. Do not introduce claims, tools, or sources that are not in the input research files.
2. Preserve caveats and disagreements. Never smooth away uncertainty.
3. Be explicit about gaps. If the research files have unresolved questions or conflicting evidence, surface them — do not paper them over.
4. Do not promote draft text into fact. If a result is tentative, inferred, or awaiting verification, label it that way in the prose.
5. No aesthetic laundering. Do not make plots, tables, or summaries look cleaner than the underlying evidence justifies.

Before finishing, the Writer does a claim sweep: every strong factual statement in the draft should have an obvious source home in the research files.

The Writer does not add inline citations or a Sources section. That is handled as a separate post-processing step by the Verifier agent. Drafts that appear to have citations added by the Writer will be treated as unverified.

Manual invocation

You can run the Writer directly on a specific task:

/run writer draft an executive summary from outputs/moe-research-papers.md

/run writer produce a paper-style draft from the scaling-laws research files

The Writer saves its output to the path specified by the lead agent (or draft.md as a fallback).