evals
by aaron-pak · author: Aaron Pak
evals is an Agent Plugin published by aaron-pak. It packages 4 skills. We fetched the plugin.json from GitHub on 2026-11-13 and checked it against the official Agent Plugins 1.0.0 schema.
Build, run and hillclimb evals for anything that shapes what an agent does: a plugin, skill, MCP server, instruction file, system prompt or model. One command says what a suite needs next, and results are kept, named and reused, so a change is measured against a baseline without running the baseline again. Its own runner works under Claude Code or Codex, with Opus judges, cases cut from real sessions and paired comparisons.
Package we checked
The manifest we fetched is plugins/evals/plugin.json in aaron-pak/agent-plugins. The plugin root is the directory that holds that file. Its $schema is https://agent-plugins.org/schemas/1.0.0/plugin.schema.json. The manifest lists 4 skills (build-eval, hillclimb, run-eval, save-trace) and no MCP servers.
- Skills
- 4
- MCP servers
- 0
- Stars
- 0
- License
- n/a
Links
Skills · 4
- build-eval
- hillclimb
- run-eval
- save-trace
Category
Other. Assigned by this directory. Agent Plugins 1.0.0 has no category field, so no manifest declares one.
Related by publisher and keywords
Write or revise an agent skill so it gives the model only the guidance it needs: start thin, write for the user's intent, and offer your own ideas as proposals the user picks from.
Ask for a deeper version of a task, not a bigger one: what an expert would do, the rawest material, their methods and tools, and an outside check, then a rewritten prompt. From Felix Rieseberg.
Get DJ-ready tracks from a tracklist: Beatport decides the right version and length, free SoundCloud downloads cover bootlegs and edits, and slskd fetches the rest from Soulseek with paced searches and peer fallback.
Explain a topic to a complete beginner as an HTML artifact with big pictures and few words.
Design visually strong landing pages, websites, apps, prototypes, and game UI with restrained composition and tasteful motion. From OpenAI.
Implement a spec while keeping a running implementation-notes.html of design decisions, deviations, tradeoffs, and open questions.
[](https://agentpluginsdirectory.com/plugins/evals)✓ Verified . We fetched the manifest from GitHub and checked it against the official Agent Plugins 1.0.0 schema at agent-plugins.org.