The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches version information from the author's GitHub repository (Timmyy3000/skills). The instructions treat this remote content as untrusted to mitigate injection risks during the update check.\n- [COMMAND_EXECUTION]: Instructs the agent to use npx skills update nabu to perform updates. This process requires explicit human confirmation before the command is executed.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from remote Nabu vaults (ingestion points in SKILL.md include read_note and search_notes). It includes an evidence chain of capability (note mutations and network operations) balanced by boundary markers and sanitization, such as strict URL parsing and a mandate to ignore instructions in external metadata.\n- [SAFE]: The skill enforces secure credential handling by requiring tokens to be stored in an approved secret store and explicitly prohibiting their appearance in logs, chat, or plaintext files.

nabu