We show what can be supported by source metadata, validation results and measured catalog activity. Missing evidence is not invented.
Compatibility labels
Native
Published in the original format used by a supported agent.
Converted
Generated from a normalized skill using a target-specific adapter.
Tested
Passed the current structural and behavioral checks for the target format.
Skill Score
New listings receive a bounded baseline score. The score can then change using measured activity such as views, installations and user ratings. Popularity is capped so it cannot fully replace quality evidence.
- Views provide a small discovery signal.
- Downloads or installations provide a stronger usage signal.
- Ratings influence the score only after real feedback exists.
- Scores remain bounded and are recalculated from stored evidence.
Security and permissions
Risk labels summarize declared permissions and available source signals. They are not a substitute for a security audit. Users should inspect scripts, network access, filesystem operations and required secrets before installation.
Corrections and updates
Authors and users can report inaccurate attribution, compatibility or licensing information through the contact page. Material corrections update the listing and its evidence.