The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill's primary function is to generate and execute new logic. In Step 4, it creates Python scripts in a functions/ directory and markdown instructions in SKILL.md. In Step 5, it explicitly directs the agent to execute these generated scripts and instructions to verify their functionality.
[INDIRECT_PROMPT_INJECTION]: The skill's analysis phase (Step 2) involves reading untrusted data from the local codebase to determine the logic and structure of the skills it generates. There is a risk that malicious content within the analyzed files could be incorporated into the generated skills.
Ingestion points: Files matching user-specified patterns across the codebase (via Grep, Glob, and Task agent).
Boundary markers: No explicit instructions for delimiters or warnings to ignore instructions inside the analyzed data.
Capability inventory: The skill uses Read, Write, Edit, Grep, Glob, Bash, and Task tools to create and test new skills.
Sanitization: No sanitization of the analyzed code patterns is performed before they are used to populate templates or scripts.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform verification steps, such as checking for the existence of generated skills and executing the created Python functions in Step 5.

nav-skill-creator