paperbanana-skill
Provides Claude Code skill definitions (and an underlying Python package) to generate publication-quality academic diagrams, statistical plots, and slide decks from text or structured data, using a multi-agent pipeline with evaluation/self-critique and provider fallback across multiple LLM/VLM/image providers.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Security posture inferred from README: uses provider API keys and an interactive setup wizard; includes a plot code injection mitigation via AST-based import blocklist (os/subprocess/socket blocked). However, README does not detail secret logging/redaction, dependency scanning, or fine-grained scopes. Network calls to multiple external providers imply data exposure considerations; no explicit data retention/residency statements.
⚡ Reliability
Best When
You want consistent academic visuals quickly, and you can supply prompts/descriptions (and optionally data/PDF) where iterative evaluation/fallback can improve quality.
Avoid When
You need deterministic outputs, formal SLAs, or strict governance that disallows automated self-critique loops and multi-provider network calls.
Use Cases
- • Text-to-figure for academic diagrams (methodology, pipeline, architecture)
- • CSV/JSON-to-academic statistical plots with auto-styling
- • Markdown/text-to-presentation slide decks with selectable style presets
- • Venue-specific (NeurIPS/ICML/ACL/IEEE) diagram styling
- • Iterative refinement loops (auto/continue with feedback)
- • Generating from PDF page ranges for diagram prompts
Not For
- • Producing medical/legal imagery that requires strict clinical/regulated validation
- • Fully offline/no-external-API environments (relies on external providers)
- • Use cases requiring a stable, formal REST API contract for programmatic integration beyond the CLI/skill workflow
- • High-assurance content pipelines where autonomous generation must be strictly audited
Interface
Authentication
Authentication is provider-key based; the skill itself is described as operating through the Claude Code plugin/skill interface and the underlying Python CLI setup.
Pricing
Costs depend on selected providers and usage; README does not quantify spend or token/image limits.
Agent Metadata
Known Gotchas
- ⚠ Outputs may be marked UNREVIEWED when the critic cannot parse/evaluate; manual review may be needed.
- ⚠ Provider capability differences (e.g., Claude VLM does not support image generation per README).
- ⚠ Non-determinism across multi-provider fallback chains and iterative loops.
- ⚠ Long/slow runs depending on iterations/auto/refinement settings; tuning may be required.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for paperbanana-skill.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.