Affected Skill
jetpack-compose/migration/migrate-xml-views-to-jetpack-compose
navigation/navigation-3
performance/r8-analyzer
play/play-billing-library-version-upgrade
system/edge-to-edge
Description
Running the skills against a Skill that validates Skills against the agentskills.io specification, there are several potential violations of the specification. Particularly with the nested levels of the documents in the references folder. The specification it self says, "Keep files one level deep".
https://agentskills.io/specification#file-references
Also there is inconsistent uses of the ./references or references/ directory structures.
The system/edge-to-edge could be broken up into a references file for the examples, so that it saves on token usage, and the examples are only loaded when needed instead of loading when the body is read.
Expected Behavior
The skills to validate and pass the skill validation python script, or a Skill validator like: https://gist.github.com/kingargyle/37e279f27880e7a92253e5f7e7686720
Actual Behavior
Several violations of the report.
Steps to Reproduce
Run the skill validator from agentskills.io or create your own validator to validate the skills being contributed to the repository.
Suggested Fix
See above. Keep in mind one of the advantages to skills is token optimization. Try to follow expected best practices, like Skill bodies are 500 lines, anything else goes to a references file.
The skills as they stand currently are heavy token users and violoate several industry best practices and specification guidelines.
See:
https://platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices#progressive-disclosure-patterns
https://agentskills.io/specification#progressive-disclosure
Affected Skill
jetpack-compose/migration/migrate-xml-views-to-jetpack-compose
navigation/navigation-3
performance/r8-analyzer
play/play-billing-library-version-upgrade
system/edge-to-edge
Description
Running the skills against a Skill that validates Skills against the agentskills.io specification, there are several potential violations of the specification. Particularly with the nested levels of the documents in the references folder. The specification it self says, "Keep files one level deep".
https://agentskills.io/specification#file-references
Also there is inconsistent uses of the
./referencesorreferences/directory structures.The
system/edge-to-edgecould be broken up into a references file for the examples, so that it saves on token usage, and the examples are only loaded when needed instead of loading when the body is read.Expected Behavior
The skills to validate and pass the skill validation python script, or a Skill validator like: https://gist.github.com/kingargyle/37e279f27880e7a92253e5f7e7686720
Actual Behavior
Several violations of the report.
Steps to Reproduce
Run the skill validator from agentskills.io or create your own validator to validate the skills being contributed to the repository.
Suggested Fix
See above. Keep in mind one of the advantages to skills is token optimization. Try to follow expected best practices, like Skill bodies are 500 lines, anything else goes to a references file.
The skills as they stand currently are heavy token users and violoate several industry best practices and specification guidelines.
See:
https://platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices#progressive-disclosure-patterns
https://agentskills.io/specification#progressive-disclosure