Baseline quality assessment
Skill krzemienski/validationforge/skills/baseline-quality-assessment
Use BEFORE making any significant code change — refactor, dependency upgrade, migration, bug fix, performance optimization — to capture an immutable 'before' snapshot. After the change, compare against this baseline to prove two things: (1) what you intended to improve actually improved, (2) what you didn't intend to touch didn't regress. Prevents the classic 'I fixed X but accidentally broke Y' failure. Reach for it on phrases like 'capture baseline', 'before refactoring', 'prove no regressions', 'snapshot current state', 'measure the improvement', or before any PR whose correctness depends on showing a delta.From its SKILL.md
npx -y skills add krzemienski/validationforge --skill baseline-quality-assessmentAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
4.5 KB, as published. Nobody here has run it
Baseline Quality Assessment
Scope
Applies before any code change that modifies existing behavior: refactoring, migration,
feature addition, dependency update, performance optimization, or bug fix. Creates an
immutable "before" snapshot in e2e-evidence/baseline/ so post-change validation can
prove: (1) changes improved what they targeted, (2) nothing previously working is broken.
Without a baseline, every claim of "no regressions" is unverifiable.
When to Use
| Scenario | Baseline Required? |
|---|---|
| Refactoring | Yes — prove behavior preserved |
| Migration | Yes — prove new stack matches old |
| Feature addition | Yes — prove existing features still work |
| Dependency update | Yes — prove nothing broke |
| Performance optimization | Yes — prove functionality preserved |
| Bug fix | Yes — capture broken state, then fixed state |
| Net-new functionality (zero existing code changes) | Optional |
4-Step Baseline Process
Step 1: Create baseline directory
mkdir -p e2e-evidence/baseline/
This directory is immutable after capture. Post-change evidence goes in e2e-evidence/, not baseline/.
Step 2: Identify all existing user journeys
Use journey discovery from create-validation-plan to list every feature, endpoint, screen, or command.
Step 3: Capture evidence for each target
Use the appropriate method per platform. See references/baseline-capture-commands.md for
platform-specific capture commands and storage path conventions.
Step 4: Document the assessment
Save to e2e-evidence/baseline/assessment.md using the template in
references/regression-comparison-template.md.
Regression Detection
| Severity | Definition | Action |
|---|---|---|
| CRITICAL | Worked in baseline, fails now | Must fix before proceeding |
| HIGH | Performance degraded >20% | Should fix before shipping |
| MEDIUM | Visual change not in requirements | Evaluate if intentional |
| Not a regression | Broken in baseline too | Pre-existing bug, document but don't block |
Post-Change Comparison
After implementing changes, re-capture the same evidence and compare side-by-side.
Save comparison to e2e-evidence/regression-check.md. See references/regression-comparison-template.md
for the full template with verdict columns (NO REGRESSION / IMPROVED / REGRESSION).
Rules
- Baseline MUST be captured before any code changes — once editing starts, "before" state is gone
- Baseline evidence is immutable — never overwrite/edit/delete files in
baseline/. Re-capture tobaseline-v2/ - Post-change evidence goes in
e2e-evidence/— not inbaseline/. Keep them separate - Regressions are CRITICAL findings — fix before claiming completion
- Pre-existing bugs are not regressions — baseline proves they existed before your changes
- Performance baselines need numbers — "it felt fast" is not a baseline
Security Policy
Baseline evidence may contain production-like data. Never commit baseline directories with
PII or credentials. Use .gitignore for e2e-evidence/.
Related Skills
- create-validation-plan — Plan journeys before capturing baseline
- preflight — Ensure system is ready before baseline capture
- e2e-validate — Execute post-change validation using same capture methods
- gate-validation-discipline — Use regression check as a completion gate
References
references/baseline-capture-commands.md— Platform-specific capture commands and storage pathsreferences/regression-comparison-template.md— Full comparison template with verdict format