The Agent Skills Directory

[PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection by processing untrusted entries from local JSONL files (data/knowledge/base/{domain}.jsonl) into formatted prompts. It currently lacks explicit boundary markers or sanitization for these exports.
[COMMAND_EXECUTION]: The skill generates Python harness modules containing rules and a prompt builder. While this involves dynamic code generation, it is a primary functional feature for developers and does not involve the execution of arbitrary shell commands or remote scripts.
[SAFE]: All file operations are restricted to local paths (data/knowledge/base and data/knowledge/exports) and align with the stated purpose of the bdistill suite.

bdistill-export