Start with your file
Paste instructions or choose Markdown. Your original stays untouched.
● AGENT INSTRUCTION QUALITY
Make your agent instructions work better.
Evaluate AGENTS.md, CLAUDE.md and SKILL.md files with clear findings, a transparent rubric, and reviewable recommendations.
✓ Browser-local free scoring ✓ Original files stay intact
● ASSESSMENT / SAMPLEILLUSTRATIVE DATA
AGENTS.mdRepository instructions · sample profile
INSTRUCTION QUALITY
78/100
Illustrative score, not measured performanceB
! Missing verification stepsAdd test commands and expected outcomes.
RUBRIC v0.1.0 · SAMPLE / EVIDENCE → RECOMMENDATION → REVIEW
BUILT FOR INSTRUCTION FILES▣ AGENTS.md▣ CLAUDE.md▣ SKILL.md
01 / FREE EVALUATION
Inspect your instructions, not just your score.
Paste Markdown or choose a local file. The current deterministic assessment runs entirely in your browser.
● INSTRUCTION ANALYZER LOCAL · NO UPLOAD
02 / WORKFLOW
A review process you can follow.
Review findings
Understand scoring dimensions, missing coverage and suggested changes.
Revise and rescore
Apply changes yourself and compare results. Advanced diffs are planned.
03 / SCORING
Transparent by design.
Current rubric v0.1.0 measures instruction-file heuristics, not downstream agent performance.
25%StructureHeadings and organization
25%ActionabilityConcrete executable steps
20%VerificationTests and expected outcomes
15%SafetyPermissions and boundaries
15%MaintainabilityDocumentation and references
04 / TRUST
Your source stays yours.
Free scoring currently happens in your browser. We don't overwrite your file or automatically change your repository. The authenticated optimization workspace is still in development.
START WITH THE FILE YOU ALREADY HAVE