skills.rest
nodatall
nodatallCOMMUNITY
Explore all skills by this author →
Website𝕏@7heThreeWordsGitHub

skill-review

Compare baseline and candidate skill versions to validate behavior changes.

2|Updated Feb 17, 2026
Search Tags:#A/B testing#regression testing#skill review#merge gate#behavior comparison#pre-merge validation#Prime Directive
One-click install
npx skills add https://github.com/nodatall/primedirective --skill skill-review-nodatall
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: skill-review
Source: https://github.com/nodatall/primedirective/tree/main/skills/skill-review
Command: npx skills add https://github.com/nodatall/primedirective --skill skill-review-nodatall

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill eliminates the risk of merging unvalidated Prime Directive skill changes that introduce regressions or fail to deliver intended improvements, by comparing actual skill behavior between baseline and candidate versions instead of only reviewing written contract prose.

Core Features & Use Cases

  • A/B Behavior Testing: Runs baseline and candidate skill versions on the same realistic scenario to compare actual outputs, not just written documentation.
  • Blind Judging: Uses a fresh subagent to evaluate run artifacts without bias toward old or new versions, producing evidence-backed verdicts.
  • Merge Gate Integration: Produces a clear pass/fail result with preserved artifacts for audit, so only proven skill changes are merged. Use Case: If you modify the $deliver skill's planning logic to add new validation steps, use this skill to run the same realistic project plan scenario against the old and new versions to confirm the change improves plan quality without breaking existing activation paths or adding unnecessary friction.

Quick Start

Invoke the skill-review skill with the name of the modified Prime Directive skill and a relevant realistic test scenario to run an A/B behavior comparison and receive a merge-ready verdict.

Frequently Asked Questions about skill-review

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

FAQPage Schema
How do I run pre-merge A/B behavior validation for skill contract changes?▼

Pre-merge A/B behavior validation compares actual execution outputs between baseline and candidate skill versions using a realistic test scenario and a fresh judging subagent to produce an evidence-backed merge readiness verdict.

What is the best way to prevent regressions when modifying Prime Directive skills?▼

To prevent regressions when modifying Prime Directive skills, run A/B behavior comparison testing that executes both baseline and candidate versions on the same scenario to verify changes improve behavior without breaking activation paths.

How does blind judging work for skill behavior comparison?▼

Blind judging uses a fresh subagent to evaluate run artifacts from baseline and candidate skill versions without bias toward old or new versions, ensuring the merge gate verdict is strictly evidence-backed.

Do I need a specific test scenario to validate skill modifications before merging?▼

Yes, pre-merge validation requires a realistic test scenario that exercises the changed behavior, alongside access to both baseline and candidate skill versions, to accurately compare execution behavior.

Can I use skill review for changes to agent configurations and validator logic?▼

Yes, pre-merge A/B behavior validation applies to any skill modification impacting behavior, including changes to SKILL.md, references, agent configurations, or validator logic, by comparing actual execution outputs.

What are the limitations of reviewing only written prose for skill contract updates?▼

Reviewing only written prose fails to catch execution regressions or broken activation paths, whereas A/B behavior comparison tests actual skill outputs against realistic scenarios to ensure proven changes are merged.

Related Skills

Explore More Software Engineering Skills→
VictoriakaeyVictoriakaey

regression-testing

Compare LLM system outputs across versions to locate regression sources.

Community
Advanced
HomericIntelligenceHomericIntelligence
18

verify-pr-ready

Validate GitHub Pull Requests for CI status, approvals, and merge conflicts.

Official
Intermediate
itallstartedwithaideaitallstartedwithaidea
31

code-review

Evaluate pull request diffs against a checklist and generate severity-classified review documents.

Community
Advanced
aspi6246aspi6246
148

code-audit

Audit R or Python research scripts with a two-phase read-only workflow.

Community
Advanced