The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the local environment by executing obsidian CLI commands (such as obsidian create, obsidian search, and obsidian read) to manage files within an Obsidian vault.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface. It is designed to ingest data from untrusted project files—including READMEs, configuration files, and source code—to automatically generate or update wiki content. The skill instructions do not specify the use of boundary markers or sanitization for this external content, potentially allowing malicious instructions embedded in a project's files to influence the agent's behavior during the wiki management process.
[SAFE]: The skill performs read-only scanning of project directories to detect file patterns (e.g., package.json, pom.xml, go.mod) for project type identification. This behavior is consistent with its primary purpose of knowledge management and does not involve unauthorized data exfiltration.

personal-kb