The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to read, audit, and modify other SKILL.md files. This creates a surface where instructions within those files could potentially influence the agent's logic during the audit or refactor process.
Ingestion points: Accesses SKILL.md files located in the workspace and the .atl/skill-registry.md index file.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard instructions found within the files it is auditing.
Capability inventory: The skill can read and write local files and execute the gentle-ai CLI tool.
Sanitization: No sanitization or safety-filtering of the content retrieved from audited skills is mentioned.
[COMMAND_EXECUTION]: The skill triggers the execution of gentle-ai skill-registry refresh to update its internal index of skills. This command is a project-specific utility for synchronizing the skill metadata.

skill-improver