agentsclimarketplace

Research manuscript audit

Skill SystemicVoid/agentic-research-skills/skills/research-manuscript-audit

Six evidence-disciplined research workflow skills for coding agents. Install with npx skills, direct download, or git clone.

Install
npx -y skills add SystemicVoid/agentic-research-skills --skill research-manuscript-audit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Audit a manuscript or technical write-up before editing and produce a prioritized evidence-linked patch plan.

SKILL.md

5.4 KB, as published. Nobody here has run it

Research Manuscript Audit

Purpose

Audit a manuscript or technical write-up before editing and produce a prioritized evidence-linked patch plan.

When To Use

Use before major manuscript edits, public release, submission, internal review, or translation into another surface. This prompt owns audit and patch planning. A later refinement prompt owns rewriting.

Required Inputs

  • {{PROJECT_NAME}}
  • {{RESEARCH_QUESTION}}
  • {{DOMAIN}}
  • {{TARGET_VENUE_OR_AUDIENCE}}
  • {{CURRENT_DRAFT}}
  • {{CANONICAL_EVIDENCE}}
  • {{AS_OF_DATE}}

Optional Inputs

  • {{FIGURES_TABLES_OR_APPENDICES}}
  • {{LITERATURE_SOURCES}}
  • {{STYLE_OR_FORMAT_REQUIREMENTS}}
  • {{PRIOR_REPORTS}}
  • {{SHARED_RESEARCH_CONTRACT}} for shared definitions covering authority order, claim statuses, provenance, and supersession.
  • {{OUTPUT_LOCATION}}
  • {{AVAILABLE_TOOLS}}

Authority Order

Use {{SHARED_RESEARCH_CONTRACT}} when supplied. Otherwise apply this order by claim type:

  1. Explicit user instructions, release constraints, safety constraints, and local project conventions.
  2. Raw or near-raw project evidence for project claims.
  3. Primary external sources for external claims.
  4. Reproducible analyses and transparent secondary sources.
  5. Canonical current-state project documents and authority ledgers.
  6. Dated archival records and historical notes.
  7. Input reports, model outputs, comments, and prior narratives, treated as leads.
  8. Editorial judgment, clearly labeled.

Tool And File-Access Assumptions

Use repository, document, citation, and browsing tools when available. Produce an audit and concrete patch plan. Apply draft edits only when the user explicitly requests edits.

Handoffs

Upstream artifacts are CURRENT_DRAFT, CANONICAL_EVIDENCE, literature synthesis, report arbitration output, accepted pipeline findings, figures, tables, appendices, and prior reports.

Downstream artifacts are Claim Audit Table, Framing And Internal Coherence, Figures And Tables Audit, Citation And Reference Audit, Prioritized Fix List, Patch Plan, and Inputs For Refinement Pass.

Pass accepted current-state findings to update-docs when documentation needs dated supersession. Pass unresolved source disagreements to report-arbitration.

Claim-Support Statuses

Use one status per substantive claim:

  • Supported
  • PartiallySupported
  • ReasonableInference
  • Ambiguous
  • Overstated
  • Unsupported
  • StaleOrSuperseded
  • Unverified
  • Opinion

Process

  1. Read the draft in full.
  2. Build a claim ledger for substantive empirical, methodological, causal, novelty, limitation, and implication claims.
  3. Assign each claim exactly one support status from the shared vocabulary.
  4. Add problem flags where useful, such as NeedsCitation, InternallyInconsistent, CitationMismatch, ScopeLeak, MissingBaseline, or ReproducibilityGap.
  5. Check whether the narrative matches the evidence, including nulls, failures, uncertainty, caveats, and boundary conditions.
  6. Audit figures and tables for necessity, comparison quality, uncertainty visibility, caption discipline, and redundancy.
  7. Audit citations for relevance, support strength, freshness, and whether they substantiate the exact claim made.
  8. Audit prose and structure for repetition, weak transitions, vague mechanism language, unsupported causal claims, and mismatch between evidence strength and wording.
  9. Compare {{PRIOR_REPORTS}} as leads when supplied.
  10. Use {{STYLE_OR_FORMAT_REQUIREMENTS}} to scope the patch plan.
  11. State tool limits from {{AVAILABLE_TOOLS}} before citation or file claims.
  12. Prioritize fixes by scientific risk and effect on reader trust.

Work privately. Expose concise rationale, evidence, uncertainty, and decisions.

Output Contract

Produce a Markdown audit with these sections:

  1. Executive Verdict covering overall quality, main scientific risk, and readiness for revision, review, or release.
  2. Claim Audit Table covering claim, location if available, claim type, support status, flags, evidence, problem, and recommended fix.
  3. Framing And Internal Coherence covering narrative tensions, overreach, missing caveats, and sections that fight the evidence.
  4. Figures And Tables Audit covering keep, revise, merge, move, or remove decisions with rationale.
  5. Citation And Reference Audit covering unsupported citation uses, missing prior art, stale sources, and citation mismatches.
  6. Prose And Structure Audit covering redundancy, unclear ordering, imprecise language, and concrete repair direction.
  7. Prioritized Fix List: ordered by impact on scientific quality.
  8. Patch Plan: file-level or section-level actions if direct editing is unavailable or not requested.
  9. Inputs For Refinement Pass covering claim ledger path or table, highest-risk sections, style constraints, and claims blocked from external-facing prose.

Write to {{OUTPUT_LOCATION}} if possible. Otherwise output the full audit in the response.

Quality Gate

Before finalizing, verify:

  • The audit leads with findings.
  • Each major criticism is tied to draft text, evidence, or missing evidence.
  • The audit preserves historical record integrity.
  • Evidence problems and prose problems are separated.
  • The audit reports only edits, browsing, and citation checks that were completed.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.