Comparison Artifacts

Comparison-Artifacts is a set of agent skills for producing the analytical artifacts that software evaluation documents typically require. Given a set of researched options, it generates feature comparison tables in tick, descriptor, or hybrid formats, a weighted scoring of options against user-defined priorities with a computed ranking and sensitivity check, and a 2x2 quadrant chart rendered as Mermaid, native Typst, or SVG. Output can be written to standalone files or inserted into a target document between named marker comments, so that re-rendering replaces only the managed block.

All three skills read and write a single comparisons/<slug>/comparison.json file per evaluation, which stores options, per-feature findings with status, descriptor and source URL, scoring configuration, and quadrant positions. The renderer is a Python 3 script built on the standard library with no external dependencies, invoked through subcommands for validation, tables, scoring, and quadrant output. Unverified cells are marked as unknown rather than estimated, and ranking is calculated by the script instead of being generated by the language model.

The project is intended for researchers and analysts writing up vendor or tool selection decisions, such as choosing a CRM, and for developers who use AI coding agents and want reproducible, source-linked comparison material in Markdown or Typst documents. The skills are packaged as plain SKILL.md files, so agents that read skills folders beyond the Claude Code plugin marketplace can also use them. The repository includes an illustrative CRM evaluation example with rendered outputs and is released under the MIT license.