agentleFS
ChangChrisLiuChangChrisLiu/academic-writing-audit/skills/scientific-story-audit/SKILL.md

scientific-story-audit

Skill4 starsChanged 32 days agoMITView on GitHub

Audit or repair the scientific argument of an abstract, introduction, or manuscript when motivation, research gap, contribution, and evidence feel disconnected. Preserve coherent passages; not a default workflow for grammar-only edits.

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:

    Before repairing a section, identify the concrete question, audience, supported gap, contribution, and principal finding.

    line 36
  2. Distinctive:

    Remove repeated background, rationale, and imagined-reviewer defenses that serve no distinct function.

    line 46
  3. Distinctive:

    Prefer connected prose, including contributions, unless the requested format requires otherwise.

    line 48
  4. Distinctive:

    Check what the reader learns, why each next step follows, and which distinction deserves emphasis.

    line 50
  5. Distinctive:

    Read across a proposed deletion.

    line 62
  6. Distinctive:

    Read the prose policy when drafting or revising manuscript prose or audit explanations.

    line 76
  7. Distinctive:

    Stop when the requested connections are clear and the claim strength matches the supplied evidence.

    line 82

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Scientific Story Audit

Find the missing connection in the research argument before changing its prose. A readable story explains why the question matters, what remains difficult, how the contribution addresses that difficulty, and what the evidence establishes.

Set the boundary

Use the current manuscript and supplied evidence. Earlier editorial judgments are leads, not verdicts. Identify whether the request is an audit, a revision, or a check after editing. Return findings for an audit; perform the bounded revision when requested. With an excerpt, assess its local argument and state which broader connection cannot be checked.

Test the strongest supported account at the study's actual stage. Locate which premise or inference a consequential objection affects. An untested hypothesis, a demonstrated contradiction, and an out-of-scope possibility call for different responses. Preserve a useful question or bounded finding; narrow or withhold an unsupported assertion. The number of objections is not a measure of rigor.

Recover the argument

Recover problem → unresolved difficulty → contribution → evidence → conclusion as a short reasoning aid, not a replacement introduction. Evidence may be an experiment, derivation, source analysis, or reasoned argument.

Check each connection:

  • Does the difficulty explain why the contribution is useful, rather than jumping from a topic to a method announcement?
  • Does the contribution name the scientific operation or framework at the right level, rather than reducing it to a coefficient or implementation detail?
  • Are complementary roles, such as acquisition, coverage, correction, inference, and selection, distinct where they matter?
  • Does the conclusion answer the original question within the evidence? A component's intended role is not an independently established effect.

At sentence boundaries, check the supporting premise, connector, referent, subject, population, and unit. A grammatical transition cannot supply missing evidence. Do not manufacture a literature gap or promote every implementation choice into a contribution to complete the chain.

For an investigated first-draft story, test its decisive gap against the closest prior solution and meaningful counterevidence in the available record. Separate documented limitations, plausible questions, and uninspected capabilities. Inspect what the methods and findings support; introductory gap assertions, prestige, and agent agreement do not settle the claim. A local audit does not require a new field search unless the requested claim depends on missing evidence.

For an apparently obvious gap, check prior attempts, constraints, and solutions under other names. Prior failure is an investigation priority, not an assumed fact or impossibility proof. Distinguish outcomes from their causes and missing reports from evidence that nobody tried. A claimed removal of an earlier barrier needs an evidenced change; a proposed change remains a hypothesis. Preserve useful replication or boundary questions. Keep unresolved causes and research debate in the handoff, with supported constraints and necessary scope in the manuscript.

Check the section logic

Before repairing a section, identify the concrete question, audience, supported gap, contribution, and principal finding. Ask only for a missing fact that determines the story, or leave that connection unresolved. A method name cannot supply it.

For an introduction, establish the venue, track or article type, readers, and submission priorities. Distinguish official or supplied criteria from inference. If the destination is unknown, continue general logic checks and ask before tailoring; do not invent venue expectations.

An ordinary abstract should normally use at most six sentences, connect task need, unresolved difficulty and contribution, and end in the principal supported result with its necessary scope. Check actual sentence boundaries without rewarding overloaded clauses. Follow explicit author or required venue formats, and keep an abstract provisional if results are absent. A module catalogue or a shortened method inventory cannot replace the reasoning.

For an introduction, check whether the chosen opening leads these readers to the actual problem. Trace that path through relevant progress, gap and motivation to contribution. Explain a competing route's relevant requirements and their consequence in this setting, retaining their precise scope; avoid repeated disadvantage lists or objections that interrupt the narrowing line. Motivation explains why the proposed move follows.

Trace one narrowing line across paragraphs and flag unrelated applications, reopened surveys, and method catalogues that interrupt it. A conference introduction may integrate gap and motivation compactly; a journal introduction should make the unresolved issue, consequence, and rationale explicit. Actual venue requirements govern, without a fixed length or heading rule.

Remove repeated background, rationale, and imagined-reviewer defenses that serve no distinct function. Retain the smallest definition, premise, or scope condition needed for the next inference. Name that connection before cutting or restoring it; explanation is not automatically defensive.

Prefer connected prose, including contributions, unless the requested format requires otherwise. Test generic importance language for a real premise before deleting it. Preserve a coherent author narrative and adapt these functions to structured abstracts, theoretical arguments, and interpretive studies.

Check what the reader learns, why each next step follows, and which distinction deserves emphasis. Restore supported rationale or qualification without inventing author deliberation or research history. Neither uniform paragraph templates nor forced variation create coherence.

Compare the expected and current argument

Use the subfield, design, audience, and retained claim to identify what the passage must establish, why, what it currently establishes, and the remaining gap. Separate actual reviewer requests from inferred reader needs. A coherent sequence can fail to answer its question, while a demanded experiment can exceed that question.

For a revision, compare supplied versions for both concern resolution and preservation of valid motivation, complementary roles, conditions, and interpretation. Keep supported corrections and useful compression. The earlier draft is not ground truth; without it, do not claim a before/after check.

Repair the smallest broken link

Locate the broken connection and prefer a bridge, corrected topic sentence, or local reorder. Follow the actual objects across the transition; records, annotations, training examples, fitted models, and outcomes are not interchangeable.

Read across a proposed deletion. Preserve a coherent passage verbatim when no concrete defect exists, and retain restatements serving a distinct purpose, such as a standalone abstract. Restore missing meaning rather than an entire old paragraph.

Worked example

Supplied facts: A crop-monitoring workflow combines scheduled observations, follow-up measurements after anomalous readings, and expert corrections. A scoring rule determines their training contributions. Only the complete workflow has been evaluated.

Weak framing: “We introduce a scoring equation; other components include follow-up visits and corrections.”

Repair: “We introduce a workflow that combines routine observations, targeted follow-ups, and expert corrections for crop monitoring. A scoring rule controls how these complementary records contribute during fitting.”

This restores the contribution's level without claiming that each component independently causes improvement. Keep any preceding motivation that already explains why heterogeneous observations are difficult to combine.

Prose when drafting or revising

Read the prose policy when drafting or revising manuscript prose or audit explanations. Preserve scientific meaning, valid terminology, necessary conditions, and sound author choices. Resolve labels against their operational definitions, and do not invent a mechanism or property in either the paper or the diagnosis.

Deliver and stop

For an audit, give the main argument defect, its location and consequence, plus explicit keep decisions. For revision, provide only the affected replacement or patch, a brief rationale, and material unresolved facts. Respect requested length and format.

Stop when the requested connections are clear and the claim strength matches the supplied evidence. “No change needed” is a valid outcome. Do not expand a story repair into a new study, exhaustive literature search, or wholesale stylistic rewrite.

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.