mikubaka88/ccfa-skills

ccf-paper-to-exemplar

Convert user-provided conference paper PDFs into distilled writing exemplar cards for the ccf-paper-writer skill.

View source
Original skill document

Rendered from the source repository. Headings, examples, code, tables, links, and referenced images are preserved.

Paper-to-Exemplar

Purpose

Convert one or more conference paper PDFs into distilled writing exemplar cards that the ccf-paper-writer skill can use as style references. This skill does not write papers. It produces exemplar cards that teach the writer skill how to write.

When To Use

  • User says: "use this paper as writing reference", "convert this PDF to exemplar", "add this to my writing templates"
  • User wants to build a personal library of favorite papers for writing-style imitation
  • User specifies a target venue and wants venue-matched exemplars loaded
  • After downloading a best paper, the user wants it distilled for fast reuse

Workflow

  1. Receive PDF paths and optional target venue from the user.
  2. Run scripts/convert.py to extract text and produce initial cards with [ANALYZE] placeholders.
  3. Read the extracted full text (.full.md file) to understand the paper content.
  4. Analyze the paper writing patterns:
  • Story arc: task -> gap -> root challenge -> insight -> method -> evidence -> limitation
  • Abstract moves: what information is packed into each sentence?
  • Introduction paragraph roles: what does each paragraph accomplish?
  • Method presentation: module-by-module, concept-first, or pipeline-first?
  • Evidence organization: table strategy, ablation logic, qualitative placement
  • Citation patterns: density, weaving style, author-name usage, grouping
  1. Load ../ccf-paper-writer/references/prose-quality-guardrails.md and filter the extracted pattern: preserve transferable logic, but do not turn defensive framing, Q1/C1 labels, formula dumping, number-only abstracts, third-person manuscript narration, punctuation-heavy structure, or unsupported hype into reusable advice.
  2. Fill in the [ANALYZE] placeholders with concrete, actionable writing-pattern descriptions.
  3. Save the completed card to ccf-paper-writer/references/exemplars/cards/.
  4. Register the card in ccf-paper-writer/references/exemplars/index.md under the correct venue section.
  5. If the user wants this as their default writing template, update ccf-paper-writer/references/custom-format/default-user-format.md.
  6. Inform the user the exemplar is ready and what writing patterns it provides.

Venue-Aware Routing

When the user specifies a target venue (e.g., "投CVPR", "for NeurIPS submission"):

  1. Classify the target venue into its venue family using ../ccf-paper-writer/references/venue-guides/index.md.
  2. If the user provides a PDF, distill it with the venue tag so it is categorized correctly.
  3. If the user does NOT provide a PDF but names a venue, check whether exemplar cards already exist for that venue:
  • Look in ../ccf-paper-writer/references/exemplars/index.md for venue-matched bundles.
  • If matching exemplars exist, load them and offer them to the writer skill.
  • If no matching exemplars exist, suggest: "No {venue} exemplars found. Provide a {venue} best paper PDF and I will distill it."
  1. When the writer skill is invoked for this venue, the exemplar index should route to the correct cards automatically.

Integration With ccf-paper-writer

The writer skill loads exemplars through references/exemplars/index.md. This skill populates that index. The handoff is:

text
paper-to-exemplar (distill PDFs -> create cards -> update index)
 -> ccf-paper-writer (load cards from index -> write with venue-matched style)

When the writer skill starts, it checks:

  1. Did the user specify a venue? If yes, load venue-matched cards.
  2. If no venue specified, load the user default cards from custom-format/default-user-format.md.
  3. If no default cards exist, load the most general-purpose cards (e.g., llava-4d, vggt).

Output Contract

After processing PDFs:

  1. List the generated card files and their locations.
  2. Summarize the writing patterns extracted from each paper.
  3. State which venue family each card belongs to.
  4. If cards were registered as defaults, confirm the registration.
  5. Provide next-action options: "Run ccf-paper-writer with these exemplars" or "Add more papers."
  6. State any excluded anti-patterns when the source paper used them heavily.

Reference Files

  • scripts/convert.py: PDF-to-markdown extraction and card skeleton generation.
  • ../ccf-paper-writer/references/exemplars/index.md: exemplar registry (this skill writes to it).
  • ../ccf-paper-writer/references/exemplars/cards/: card storage directory.
  • ../ccf-paper-writer/references/custom-format/default-user-format.md: default exemplar configuration.
  • ../ccf-paper-writer/references/venue-guides/index.md: venue family classification.
  • ../ccf-paper-writer/references/prose-quality-guardrails.md: anti-pattern filter for distilled writing cards.

Card Format

Every exemplar card follows this structure (see existing cards for examples):

markdown
# Paper Title

Venue/year: CVPR2025.
Source: distilled from PDF by ccf-paper-to-exemplar.
Use when: writing {paper type} in the {venue} venue family.

## Story Pattern
Describe the story arc the paper follows.

## Abstract Moves
How does the abstract compress the contribution?

## Introduction Moves
Paragraph-by-paragraph role analysis.

## Method Moves
How is the method presented and justified?

## Evidence Moves
Evidence types, table/figure strategy, claim mapping.

## Citation Patterns
How are citations woven? Density? Style?

## Reusable Techniques
Transferable writing techniques.

## Excluded Anti-Patterns
Writing moves from the source that should not be reused, such as label-heavy prose, formula dumping, number-only abstract style, defensive framing, or punctuation-driven structure.

## Do-Not-Copy Boundary
Do not copy specific task, claims, examples, or technical content.

Dependencies

  • Python3 with pymupdf (install: pip install pymupdf)
  • The convert.py script handles PDF text extraction
from this repository

More skills

All skills
mikubaka88
Community

ccf-common

Shared governance for the CCFA family: routing, trigger registry, task modes, handoff modes, source registry, privacy/evidence policy, ccfa.yaml, and artifact contracts. Use only when maintaining or auditing CCFA skills; not for ordinary research tasks.

installs
5
GitHub stars
2 k
Updated
13 août
mikubaka88
Community

ccf-experiment-designer

Own experiment evidence semantics: decide datasets, baselines, metrics, ablations, robustness tests, chart evidence, and exactly what rows or columns a result table should contain. Use for experiment design, benchmark planning, supplied-result evidence structure, result-table schema, chart-spec semantics, 设计实验, 对比实验, 消融, 结果表证据结构. Do not search literature as the main deliverable, visually beautify or render an already specified table/figure, improve layout/colors/readability, or invent results.

installs
5
GitHub stars
2 k
Updated
13 août
mikubaka88
Community

ccf-humanization

Humanize and de-defend publication-facing CCF/AI prose and artifacts as a sidecar preflight, never as the primary task owner. Use before manuscript drafting/revision and before final manuscript experiment text, tables, captions, or method descriptions when Codex must remove defensive writing, boilerplate disclaimers, repetitive improbable edge cases, unnecessary safeguards, generic SHA-256/checksum requirements, duplicated smoke tests, simplified/toy methods, or AI-like risk narration. Do not auto-load as the sole owner for raw experiment planning, retrieval, review, auditing, routing, visual rendering, or assessment-only tasks without publication prose. Keep necessary concerns in a separate user-review warning instead of silently injecting them into files. Do not conceal material evidence, fabricate results, or override venue-mandated disclosures.

installs
5
GitHub stars
2 k
Updated
13 août
mikubaka88
Community

ccf-idea-optimizer

Turn rough CCF research directions into concrete problem, gap, insight, method, novelty, and evidence plans. Use for idea optimization, fuzzy idea concretization, develop a fuzzy idea with no score, research direction shaping, early direction exploration, salvage routes, 优化idea, 具象化idea, 研究思路优化, 找方向, 方向探索. Do not rank or score multiple ideas as the main output.

installs
5
GitHub stars
2 k
Updated
13 août