agentleFS
aidlessaidless/research-writing-skill/skills/evidence-driven-writing/SKILL.md

evidence-driven-writing

Skill0 starsChanged 40 days agoMITView on GitHub

Use when writing or revising Introduction, Related Work, background, literature synthesis, or any section where references must drive claims

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not write the section until an evidence map and paragraph blueprint exist.

    line 12
  2. Distinctive:

    Create one row per usable source:

    line 22
  3. Distinctive:

    Use the user's literature pool first.

    line 29
  4. Distinctive:

    Use only information present in title, abstract, DOI metadata, or user-provided notes unless full text is available.

    line 30
  5. Distinctive:

    Only after the blueprint is complete may manuscript prose be written.

    line 47
  6. Distinctive:

    Do not end the Introduction with a long mechanical chapter map.

    line 61
  7. Distinctive:

    If a paragraph can be converted into a table row without losing meaning, it is not yet manuscript prose.

    line 88

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Evidence-Driven Writing

This skill turns a literature pool into manuscript argument. It is required for Introduction, Related Work, and any background section with citations.

Hard Gate

Do not write the section until an evidence map and paragraph blueprint exist.

Required artifacts:

  • refs/evidence-map.md or plan/evidence-map.md
  • plan/chapter-blueprints/<section>-blueprint.md
  • plan/review/evidence-coverage.md

Evidence Map

Create one row per usable source:

Source IDCitationSource typeAbstract-level findingUsable factSupported claimCitation slotRisk

Rules:

  • Use the user's literature pool first.
  • Use only information present in title, abstract, DOI metadata, or user-provided notes unless full text is available.
  • One citation must support one concrete claim, not a vague field statement.
  • Mark weak or indirect support as Risk: indirect.

Paragraph Blueprint

For each paragraph, define:

### Paragraph N
- Role: context / method landscape / limitation / gap / contribution
- Main claim:
- Evidence IDs:
- Contrast or transition:
- Forbidden content:

Only after the blueprint is complete may manuscript prose be written.

Introduction Pattern

A publishable technical Introduction usually follows this chain:

  1. Application context and why the problem matters, supported by foundational or survey citations.
  2. Existing technical routes, grouped by method family rather than by paper list.
  3. Bottleneck cascade: what each route solves and what remains unsolved.
  4. Specific research gap that the proposed method can realistically address.
  5. Contributions written as prose or a short list only if the target style allows it.

For computer science / engineering SCI drafts, integrate the most relevant related-work synthesis into the Introduction unless the locked outline explicitly requires a standalone Related Work chapter. The section must read as an argument: field pressure, existing routes, unresolved bottleneck, proposed position, and contribution boundary.

Do not end the Introduction with a long mechanical chapter map. One concise manuscript-organization paragraph is enough.

Related Work Pattern

Organize by theme, not by chronological paper dump:

  • Theme opening: define the method family or research stream.
  • Evidence synthesis: compare 2-4 sources in the same paragraph.
  • Critical boundary: state what the theme does not solve.
  • Bridge: explain why the next theme or the proposed method is needed.

Body Contamination Firewall

User requirements and process notes must not enter manuscript text. Phrases such as "write naturally", "avoid generic wording", "discussion prompt", "fill later", "user should replace", and "this section is a template" belong in plan files, never in chapters.

Compression is not deletion. When shortening, preserve the claim, evidence, method condition, and limitation.

Anti-Enumeration Gate

Literature-driven sections fail review if they read like a sequence of source notes. Each paragraph must synthesize at least two of the following:

  • a problem condition;
  • a method family;
  • a limitation shared by multiple studies;
  • a direct bridge to the proposed method;
  • a citation-backed boundary.

If a paragraph can be converted into a table row without losing meaning, it is not yet manuscript prose.

GPR Sample

For a reusable example of literature-driven argument structure extracted from the user's GPR passage, read references/gpr-introduction-example.md. Use it as a pattern, not as content for unrelated papers.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.