AgentSkillsCN

review-grant

完整的资助评审流程——对T32申请书依次执行合规性检查、研究部分评审以及校对工作,并将所有结果汇总成一份综合报告。

SKILL.md
--- frontmatter
name: review-grant
description: Full grant review pipeline — runs compliance checker, study section reviewer, and proofreader on the T32 proposal. Aggregates all results into a single summary report.
disable-model-invocation: true
argument-hint: "[optional: compliance | review | proofread — run only one stage]"

Full Grant Review

Orchestrate a comprehensive review of the T32 grant proposal.

Steps

  1. Parse the argument:

    • $ARGUMENTS = empty or all → run all three stages
    • $ARGUMENTS = compliance → run only compliance checker
    • $ARGUMENTS = review → run only study section reviewer
    • $ARGUMENTS = proofread → run only proofreader on all .tex files
  2. Stage 1: Compliance Check Run the t32-compliance-checker agent on the proposal.

    • This checks section completeness, page limits, required content elements, and formatting
    • Output: quality_reports/t32_compliance_report.md
  3. Stage 2: Study Section Review Run the nih-study-section-reviewer agent on the proposal.

    • This simulates a T32 study section review with scores and critique
    • Output: quality_reports/study_section_review.md
  4. Stage 3: Proofreading Run the proofreader agent on all .tex files in proposal/.

    • Check each section file for grammar, typos, consistency, and academic quality
    • Output: quality_reports/proposal_proofread_report.md
  5. Aggregate Results After all stages complete, create a summary report at quality_reports/grant_review_summary.md:

    markdown
    # Grant Review Summary
    
    **Date:** [YYYY-MM-DD]
    **Stages Run:** [list]
    
    ## Compliance
    - **Status:** PASS / FAIL / PASS WITH WARNINGS
    - **Critical Issues:** [count and summary]
    - **Warnings:** [count]
    
    ## Study Section Review
    - **Overall Impact Score:** [N]/9
    - **Criterion Scores:** [list each]
    - **Top Strengths:** [top 3]
    - **Top Weaknesses:** [top 3]
    
    ## Proofreading
    - **Issues Found:** [count by severity]
    - **High Priority:** [list]
    
    ## Recommended Actions (Priority Order)
    1. [Most critical action]
    2. [Next action]
    3. ...
    
  6. Report to user:

    • Summary of findings across all three stages
    • Paths to all generated reports
    • Recommended next steps in priority order