Run local LLM-agent regression reviews against a baseline with mock test cases.
npx skills add https://github.com/mr-kelly/skills --skill kelly-agent-eval
Please help me install this Agent Skill. Skill: kelly-agent-eval Source: https://github.com/mr-kelly/skills/tree/main/skills/kelly-agent-eval Command: npx skills add https://github.com/mr-kelly/skills --skill kelly-agent-eval
This skill solves the challenge of maintaining quality in LLM-agent workflows by providing a structured, deterministic way to catch regressions before a release. It prevents quality drift by comparing new agent versions against a established baseline using a fixed suite of test cases.
Use the kelly-agent-eval skill to open the local review board and triage the current candidate agent regressions.
High-intent search queries and answers about installing and using this skill.
Testing agent performance regressions requires running a fixed suite of mock test cases against a candidate version. It compares results against a baseline version using a four-part rubric to identify quality drift in support, reasoning, and safety categories.
A local review board manages human-in-the-loop release decisions by providing a dashboard to triage detected regressions, record human decisions, and approve or block releases based on deterministic test outcomes.
Yes, agent regression testing with Hono requires a local environment to execute the server. It relies on local execution to run the Hono-based server and manage handoff files for release reporting.
The best way to evaluate agent quality drift is to compare new agent versions against an established baseline. This approach automatically identifies cases where a candidate agent scores lower on a four-part rubric.
You should use a baseline comparison for agent quality assurance before deploying a new version. This verifies that the new model maintains performance across categories without introducing new errors.
jbaranskiInstall and configure the latest Node.js LTS version using nvm.
DhumitechPlan and execute safe deployment rollouts with blue-green, canary, and rollback procedures.
jamesrochabrunGenerate QA test plans, manual cases, regression suites, and bug reports.
DhumitechAutomate AI agent evaluation and benchmarking with repeatable test pipelines.
