Analyze papers, GitHub repositories, documentation, notebooks, benchmarks, and AI roadmaps with one reusable Knowledge Compilation Framework for Codex.
Paper Companion Writer reconstructs complex technical material into accurate, teachable, publication-ready assets. Instead of preserving the source's original structure, it extracts concepts and evidence, rebuilds their dependencies, and packages the result for learning, engineering review, or publication.
Generate only the assets your task needs:
- Executive summaries
- Critical and engineering reviews
- Knowledge graphs and concept trees
- Learning paths and milestone timelines
- Repository and architecture analyses
- Notebook and experiment critiques
- Benchmark comparisons
- Challenge exercises and portfolio projects
- Premium deep dives
- Publication-ready Markdown
- Reference collections and visual asset plans
| Traditional Summarizer | Paper Companion Writer |
|---|---|
| Produces a summary | Reconstructs reusable knowledge |
| Follows one document | Analyzes single or mixed sources |
| Preserves source order | Reorders concepts by learning dependency |
| Offers limited critique | Evaluates evidence and engineering trade-offs |
| Produces one static response | Generates multiple reusable artifacts |
| Stops after drafting | Applies checklists, scoring, and quality gates |
| Requires manual repackaging | Produces publication-ready outputs |
flowchart TD
A["Technical Source"] --> B["Content Detection"]
B --> C["Knowledge Extraction"]
C --> D["Knowledge Reconstruction"]
D --> E["Learning Asset Generation"]
E --> F["Publication Pipeline"]
F --> G["Quality Validation"]
The framework separates source facts, observed evidence, interpretation, and recommendations. It then reorganizes the validated knowledge around what readers need to understand first.
A research paper describing a cache-aware retrieval method,
its experiment, reported results, and limitations.
Use $paper-companion-writer to analyze this paper for intermediate
software engineers. Produce an executive summary, critical review,
learning path, and knowledge graph in Traditional Chinese.
outputs/
├── summary.md
├── critical-review.md
├── learning-path.md
└── knowledge-graph.md
Each artifact shares one evidence base while serving a different learning or publication purpose.
Invoke $paper-companion-writer with four pieces of context:
- Source — a file, URL, repository, notebook, or mixed source set
- Audience — beginner, intermediate, advanced, or a specific role
- Deliverables — the assets you want generated
- Output requirements — language, format, platform, and destination
Use $paper-companion-writer to analyze this repository for intermediate
Python engineers. Generate a repository analysis, learning path, and
review report in Traditional Chinese. Write them to outputs/repository/.
The skill detects the source type, loads the matching workflow, reconstructs the knowledge, renders the requested templates, and applies the relevant quality gates.
| Source | Analysis Focus |
|---|---|
| Research papers | Motivation, contribution, method, evidence, limitations |
| GitHub and local repositories | Architecture, execution flow, engineering quality, learning value |
| Technical documentation | Concepts, API workflows, developer experience, missing knowledge |
| Notebooks and experiments | Reproducibility, data and model pipelines, experiment quality |
| Benchmarks | Evaluation design, metric validity, fairness, engineering impact |
| AI and learning roadmaps | Prerequisites, dependency order, projects, career readiness |
| Books, videos, and technical blogs | Knowledge extraction and educational reconstruction |
| Mixed resources | Independent analysis followed by evidence-aware synthesis |
Paper Companion Writer v3.0.0 passed a complete release audit rather than relying on a production-readiness claim alone.
- ✅ Zero dependency cycles
- ✅ Zero broken internal references
- ✅ Complete prompt-to-workflow routing
- ✅ All templates reachable
- ✅ Scoring weights validated at 100%
- ✅ Source and distribution packages validated
- ✅ Git object integrity checks passed
- ✅ MIT licensed
- ✅ General Availability release
See the audit report and release report for the supporting evidence.
The user-facing pipeline is implemented through seven explicit framework layers:
references → workflow → prompts → templates → checklists → scoring → automation
| Layer | Responsibility |
|---|---|
references/ |
Canonical rules and source-analysis guides |
workflow/ |
Detection, analysis, reconstruction, review, and publication contracts |
prompts/ |
Executable instructions that load the correct workflow |
templates/ |
Reusable output structures |
checklists/ |
Technical, educational, and publication quality gates |
scoring/ |
Evidence-based evaluation rubrics |
automation/ |
Lint → validate → score → export pipeline |
Start with SKILL.md for routing and execution instructions.
- Understand before explaining.
- Structure before details.
- Evidence before judgment.
- Learning order before source order.
- Generate reusable knowledge once; adapt it everywhere.
- Never invent claims, measurements, citations, or implementation behavior.
Paper Companion Writer v3.0.0 is available as a General Availability release under the MIT License. The self-contained package is available in dist/.