Judgment Skills

Agent Skills that teach AI coding assistants how to work with Judgment for tracing, evaluations, code judges, datasets, MCP server workflows, and agent performance workflows.

Skills

Skill	Description
judgment	Main skill for adding Judgment tracing and evaluations, choosing scorer patterns, using code judges, and following Judgment SDK best practices.
mcp-server-best-practices	Standalone skill for using the Judgment MCP server effectively, including full-text trace search, batched queries, and production data workflows.
judgeval-jql	Query and analyze project traces, spans, and sessions with tenant-safe JQL, including aggregations, pipelines, presentations, and discovery.

Installation

Use your coding agent

Use your coding agent with this instruction so it can install the Judgment skill and apply it to your task.

Install the Judgment skill from github.com/JudgmentLabs/skills
and use it to add tracing to this application
following Judgment best practices.

For MCP-specific workflows:

Install the mcp-server-best-practices skill from github.com/JudgmentLabs/skills.

For public Judgeval JQL:

Install the judgeval-jql skill from github.com/JudgmentLabs/skills.
Detect whether this project uses Judgeval Python or TypeScript, then use the
matching public SDK to build tenant-safe JQL for: <request>.

Cursor

Install as a Cursor plugin:

/add-plugin judgment

Or via the skills CLI:

npx skills add JudgmentLabs/skills --skill "judgment" --agent cursor
npx skills add JudgmentLabs/skills --skill "mcp-server-best-practices" --agent cursor
npx skills add JudgmentLabs/skills --skill "judgeval-jql" --agent cursor

Claude Code

Add the marketplace and install:

claude plugin marketplace add JudgmentLabs/skills
claude plugin install judgment@judgment-skills

Or via the skills CLI:

npx skills add JudgmentLabs/skills --skill "judgment" --agent claude-code
npx skills add JudgmentLabs/skills --skill "mcp-server-best-practices" --agent claude-code
npx skills add JudgmentLabs/skills --skill "judgeval-jql" --agent claude-code

Install with npx

npx skills add JudgmentLabs/skills --skill "judgment"
npx skills add JudgmentLabs/skills --skill "mcp-server-best-practices"
npx skills add JudgmentLabs/skills --skill "judgeval-jql"

Direct Judgeval JQL references

Agents that cannot install a skill can fetch the stable docs-hosted files directly:

• Concise workflow
• Python reference
• TypeScript reference
• Complete one-fetch bundle

The immutable JQL source commit and SHA-256 hashes for this publication are recorded in source.json and verified in CI.

Prerequisites

Set your Judgment credentials before asking an agent to run traces or evaluations:

export JUDGMENT_API_KEY=...
export JUDGMENT_ORG_ID=...

Usage

Once installed, your agent can use these skills when you ask it to:

• Add Judgment tracing to an agent or workflow
• Test model, prompt, tool, or agent config changes with OfflineTracer before production
• Evaluate agent outputs with Judgment
• Create or debug Python code judges
• Choose scorer patterns for offline and hosted evaluations
• Use the Judgment MCP server for production traces, behaviors, prompts, automations, agent memory, and agent threads
• Query project traces, spans, and sessions with tenant-safe Judgeval JQL
• Look up current Judgment docs and SDK reference pages