The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from session logs and configuration files which are external and potentially untrusted. Ingestion points: session logs located at ~/.claude/projects/ and configuration files such as CLAUDE.md, AGENTS.md, and CODING_STANDARDS.md. Boundary markers: None explicitly defined for the interpolation of log data into the prompt, though the process requires user review of suggestions. Capability inventory: Suggests updates to environment configuration files and uses the create-skill tool. Sanitization: The skill contains explicit instructions to redact secrets when quoting from session logs.
[COMMAND_EXECUTION]: The skill references a vendor-specific command /hb:architecture-audit and the create-skill tool for environment optimization. These are used to generate suggestions that must be manually approved by the user.

retro