OpenAPI Spec Quality
The API reference pages are only as good as the OpenAPI specs behind them. The microservice-api-docs plugin scores each spec, blocks sensitive data, and fixes gaps in the service's source code — not in the generated spec.
Scoring
The quality skill reads the generated specs in docs/specs/, together with docs/intro.md and docs/environments.json. It never reads the service's source code and changes no files, so it works on any stack.
| Metric | Weight |
|---|---|
| Operation descriptions | 25% |
| Parameter descriptions | 20% |
| Request examples | 15% |
| Response examples | 15% |
intro.md quality | 15% |
| Schema completeness | 10% |
A spec passes with a score of 60 or above and no blockers.
Sensitive Data
Sensitive data is a blocker — a spec that contains it fails, whatever its score. The skill checks for:
- Specific partner IDs, especially negative (internal) ones
- Internal hostnames, private IP addresses, and internal-only URLs
- Credentials, secrets, and infrastructure details
Only public *.kaltura.ai and *.kaltura.com URLs can appear in published docs.
Fixing Low Scores
If the score is below the threshold, the plugin recommends the fix skill that matches the repository's technology:
| Skill | Stack | What it changes |
|---|---|---|
fix-nestjs | NestJS + @nestjs/swagger | Controllers and DTOs |
fix-fastapi | FastAPI + Pydantic | Routes and models |
The fix skills add missing operation summaries, property descriptions, and request and response examples. They only add what is missing and follow the repo's existing conventions, so running them again on a fixed repo changes nothing.
Running It Yourself
Run these from the root of the service repo:
/microservice-api-docs:quality # score and list gaps
/microservice-api-docs:fix-nestjs # NestJS: fix, regenerate, re-score
/microservice-api-docs:fix-fastapi # FastAPI: fix, regenerate, re-score
Missing docs/intro.md or docs/environments.json? Run /microservice-api-docs:init to generate starter versions from the spec.
The same skills also run automatically on every staging deploy — see Staging & Production.
To change what the fix skills do or which data counts as sensitive, see Updating the Quality Skills.