internal-eval

Coordinate capture, run, score, and validate sub-skills into an end-to-end evaluation workflow.

1|Updated Mar 15, 2026

Search Tags:#isolation #evaluation #orchestration #diff #tests #harness

Ready to run

npx skills add https://github.com/paulingham/.claude --skill internal-eval

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: internal-eval
Source: https://github.com/paulingham/.claude/tree/main/skills/internal-eval
Command: npx skills add https://github.com/paulingham/.claude --skill internal-eval

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Coordinates the end-to-end evaluation of harness runs by orchestrating intake, planning, building, reviewing, final verification, shipping, deploying learnings, and recording verdicts to gate changes.

Core Features & Use Cases

Orchestrates the full evaluation pipeline across multiple phases and sub-skills (capture, run, score, and validate) with deterministic verdicts.
Provides baseline capture, regression scoring, and audit-friendly reports to track improvements and regressions.
Enforces isolation and governance rules to prevent leakage between outer and inner pipelines during nested runs.

Quick Start

Run the top-level run-suite.sh to execute the full evaluation pipeline across all cases.

Frequently Asked Questions about internal-eval

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I orchestrate end-to-end harness evaluation for regression testing?▼

To orchestrate end-to-end harness evaluation, you can coordinate capture, run, score, and validate sub-skills to produce deterministic per-run verdicts and regression reports. This workflow gates harness changes consistently across privacy, baseline capture, and diffing tasks.

What is test harness isolation and why is it needed during nested runs?▼

Test harness isolation separates inner pipelines from outer orchestration during nested runs to prevent state leakage. It enforces governance rules across evaluation phases, ensuring that baseline capture and regression scoring remain accurate and deterministic.

How do I generate audit-friendly regression reports from test runs?▼

You generate audit-friendly regression reports by running the full evaluation pipeline, which coordinates intake, scoring, and validation. It writes structured results for downstream tooling, tracking improvements and regressions to gate harness changes.

Can I use this evaluation orchestration workflow without any external dependencies?▼

Yes, you can use this evaluation orchestration workflow without external dependencies. It operates standalone to coordinate intake, planning, building, reviewing, and shipping phases, producing structured verdicts to gate changes across privacy and diffing tasks.

How do I run the full evaluation pipeline across all test cases?▼

You run the full evaluation pipeline across all cases by executing the top-level run-suite.sh script. This coordinates the end-to-end workflow across multiple phases and sub-skills, producing per-run verdicts and deploying learnings.