The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands via the Bash tool to scan the project directory and identify technology stacks and patterns.
[INDIRECT_PROMPT_INJECTION]: The skill derives instructions for newly generated skills from project-local data, which could be influenced by malicious content in the scanned files.
Ingestion points: The skill reads CLAUDE.md, .bkit/btw-suggestions.json, and various project source files.
Boundary markers: No specific delimiters or safety instructions are used to isolate untrusted content during the skill generation process.
Capability inventory: The skill leverages Bash, Write, and Edit tools to perform its analysis and file creation.
Sanitization: File content ingested during analysis is not sanitized before being incorporated into generated skill instructions.

skill-create