The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's core functionality relies on the getnote CLI tool. The instructions in SKILL.md and the references/ folder provide a rigorous framework for executing commands, including mandatory use of -o json for machine-readable output, handling of exit codes, and safe input methods (stdin/files) for long content to prevent shell injection or data corruption.
[INDIRECT_PROMPT_INJECTION]: Because the skill processes external content (web links, OCR from images, and user notes), it possesses an indirect prompt injection surface. The documentation includes specific mitigations, instructing the agent to treat retrieved content as data rather than instructions and requiring explicit confirmation for sensitive actions like deleting or sharing notes.
[EXTERNAL_DOWNLOADS]: The skill declares its dependency on the @getnote/cli NPM package through the platform-managed install field in the SKILL.md frontmatter. This allows the host platform to oversee the installation process rather than the skill executing unverified download scripts at runtime.
[DATA_EXPOSURE]: The skill is designed to handle private user notes. It includes explicit privacy guidelines, such as restricting the expansion of private notes in shared sessions and prohibiting the agent from requesting or displaying sensitive authentication credentials like API keys, tokens, or cookies.

getnote