ToonDex creates semantic folder indexes for LLM codebase navigation. It generates an index.toon file with concise summaries of each folder, ordered by importance.
93
Quality
93%
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Combined evidence-based and self-assessment scoring for INDEX.toon summaries.
| Factor | Points | How to Verify |
|---|---|---|
| README or docs exist in folder | +2 | Check for README.md, docs/, CONTRIBUTING.md |
| Read 3+ source files | +2 | Actually opened and analyzed files |
| Recognized framework/pattern | +1 | Identified Next.js, React, Express, etc. |
| Clear single responsibility | +1 | Folder has one obvious purpose |
| Certainty Level | Points | Criteria |
|---|---|---|
| High | +4 | Clear purpose, read docs, confident in summary |
| Moderate | +2 | Reasonable inference, some uncertainty |
| Low | +0 | Guessing from names/structure only |
confidence = (evidence_points + self_assessment_points) / 10| Score Range | Label | Action |
|---|---|---|
| ≥ 0.7 | High | No review needed |
| 0.4 - 0.7 | Medium | Optional review |
| < 0.4 | Low | User review required |
Folder: src/auth
Evidence: README exists (+2), read 5 files (+2), Clerk pattern (+1), single purpose (+1) = 6
Self-assessment: High certainty (+4)
Total: 10/10 = 1.0Folder: src/utils
Evidence: No README (+0), read 3 files (+2), no clear pattern (+0), single purpose (+1) = 3
Self-assessment: Moderate (+2)
Total: 5/10 = 0.5Folder: src/legacy
Evidence: No README (+0), couldn't read files (+0), unknown pattern (+0), unclear purpose (+0) = 0
Self-assessment: Low (+0)
Total: 0/10 = 0.0When analyzing folders, report confidence inline:
src/auth,Clerk-based authentication. Role middleware protects /dashboard/*. [0.9]
src/utils,Helper functions for dates and strings. [0.5]
src/legacy,Unknown purpose. Contains old code. [0.2]The bracketed score is stripped when writing final INDEX.toon but used to determine which items need user review.
Install with Tessl CLI
npx tessl i cpoepke/toon-dex@0.3.0