AgentSkillsCN

abt-narrative-critique

采用 Olson 的《Houston, We Have a Narrative》中的 And-But-Therefore (ABT) 叙事框架，对科研项目申请书和论文进行评估。该框架专为环境科学与城市气候研究领域优化设计。适用场景如下：(1) 用户希望对科研项目申请书的引言或摘要进行评析与优化；(2) 用户希望评估论文框架是否具备发表吸引力；(3) 用户需要关于科学写作叙事结构的反馈；(4) 用户提到 ABT 框架、Olson 或《Houston We Have a Narrative》一书。本技能将提供定性评析，并对问题的严重程度进行分级（致命缺陷 vs. 可改进问题），同时进行引用核查，并提出建设性改进建议。

SKILL.md
--- frontmatter
name: abt-narrative-critique
description: "Evaluate research proposals and papers using the And-But-Therefore (ABT) storytelling framework from Olson's 'Houston, We Have a Narrative'. Calibrated for environmental science and urban climate research. Use when: (1) User asks to critique or evaluate a research proposal introduction/abstract, (2) User asks to assess whether a paper's framing is compelling for publication, (3) User wants feedback on narrative structure of scientific writing, (4) User mentions ABT framework, Olson, or 'Houston We Have a Narrative'. Provides qualitative critique with severity assessment (fatal vs. improvable), citation verification, and constructive suggestions."

ABT Narrative Critique

Evaluate scientific writing using the And-But-Therefore framework. Focuses on introductions and abstracts of research proposals and papers in environmental/urban science.

The ABT Framework

Scientific narratives follow: AND (context) → BUT (tension) → THEREFORE (resolution)

  • •AND: Establishes what we know. Cites relevant literature to frame the problem.
  • •BUT: Identifies the critical gap. Creates tension that demands resolution.
  • •THEREFORE: Proposes the solution. Shows how methods/results address the gap.

Workflow

1. Determine Document Type

  • •Research proposal: Forward-looking; "Therefore" focuses on proposed methods and expected contributions
  • •Research paper: Retrospective; "Therefore" includes results and demonstrated impact

See references/proposal-criteria.md or references/paper-criteria.md for type-specific evaluation criteria.

2. Extract ABT Components

Parse the introduction/abstract to identify:

  • •AND section: Background statements, cited literature, established knowledge
  • •BUT section: Gap statement, tension, "however/but/yet" pivot
  • •THEREFORE section: Proposed approach, methods justification, expected/demonstrated outcomes

3. Evaluate Each Component

AND Evaluation

  1. •Citation coverage: Extract all cited works. Search for each using Hugging Face paper search and web search to verify:
    • •Citations exist and are from credible sources (peer-reviewed journals, reputable institutions)
    • •Cited claims accurately represent source content (no distortion)
    • •Coverage is appropriate—neither too sparse (missing key works) nor excessive (unfocused)
  2. •Framing quality: Do citations build a coherent foundation for the research question?

BUT Evaluation

  1. •Gap clarity: Is there a clear, specific knowledge gap?
  2. •Gap significance: Why does this gap matter? Consequences for:
    • •Scientific field advancement
    • •Technology development
    • •Policy interventions
    • •Societal problems
  3. •Logical connection: Does the gap emerge naturally from the AND section?

THEREFORE Evaluation

  1. •Method-gap alignment: Is there a "glove-to-hand" fit between problem and approach?
  2. •Feasibility/novelty: Are methods appropriate and sufficiently innovative?
  3. •For papers only: Do results actually address the stated gap? What are broader implications?

4. Verify Citations (Full Retrieval)

For each citation in the AND section:

code
1. Search: Use Hugging Face:paper_search and web_search to locate the paper
2. Fetch: Use web_fetch to retrieve abstract/full text when available
3. Verify: Check that the cited claim matches the source's actual findings
4. Assess: Is this source credible? Peer-reviewed? Recent enough?

Flag issues:

  • •Citation not found or inaccessible
  • •Claim misrepresents source
  • •Source is non-peer-reviewed or low credibility
  • •Key recent literature missing

5. Assess Severity

Classify each issue as:

  • •Fatal flaw: Likely to cause rejection; must be addressed (e.g., missing gap statement, methods don't address stated problem, citation misrepresents source)
  • •Significant weakness: Reduces competitiveness but not disqualifying (e.g., gap significance unclear, citations incomplete)
  • •Minor issue: Polish-level improvement (e.g., phrasing, flow)

6. Generate Critique

Structure output as:

code
## ABT Narrative Assessment

### Overall Verdict
[One sentence: Strong/Adequate/Weak narrative structure]

### AND (Context & Literature)
[Assessment of framing and citations]
- Severity: [Fatal/Significant/Minor or None]

### BUT (Knowledge Gap)
[Assessment of gap identification and significance]
- Severity: [Fatal/Significant/Minor or None]

### THEREFORE (Resolution)
[Assessment of method-gap fit and implications]
- Severity: [Fatal/Significant/Minor or None]

### Citation Verification
[Summary of citation checks; flag any issues]

### Key Recommendations
**High-level**: [1-2 sentences on most important improvement]

**Specific suggestions**:
1. [Concise, actionable suggestion]
2. [Concise, actionable suggestion]
3. [Concise, actionable suggestion]

Field Calibration: Environmental/Urban Science

When evaluating citation coverage, consider these domain norms:

  • •Urban climate papers typically cite 20-40 references; introductions draw on 8-15 key works
  • •Essential literature includes foundational works (Oke, Stewart & Oke LCZ) plus recent advances
  • •Interdisciplinary framing is valued—connections to public health, energy, planning strengthen the AND
  • •Methods sections should reference validation studies and data sources

Common fatal flaws in this field:

  • •Ignoring scale mismatches (e.g., claiming city-level implications from point measurements)
  • •Citing modeling studies as observational evidence
  • •Missing recent high-impact papers in rapidly evolving subfields (heat exposure, urban scaling)

Important Notes

  • •Critique only: Do not rewrite the document. Provide assessment and suggestions.
  • •Constructive tone: Frame feedback to help improve the work, not dismiss it.
  • •Acknowledge strengths: Note what works well alongside areas for improvement.
  • •Uncertainty: If unable to verify a citation, note this rather than assuming incorrectness.