The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the @geminilight/mindos package from the npm registry, which is the official CLI tool provided by the vendor for interacting with the knowledge base.
[COMMAND_EXECUTION]: The skill interacts with the user's local system using the mindos CLI tool. It performs scoped operations like file list, file read, file write, and search, ensuring that all actions are performed through a structured interface.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests and processes content from a user's knowledge base.
Ingestion points: External data enters the agent context via mindos file read and mindos search commands.
Boundary markers: The instructions do not explicitly specify delimiters or 'ignore' warnings for the data being read.
Capability inventory: The skill possesses file system capabilities including writing, deleting, and renaming files via the mindos CLI.
Sanitization: There is no mention of explicit sanitization or filtering of the content retrieved from the knowledge base files.
[SAFE]: The skill implements strong safety rules, such as prohibiting writes to the root directory, requiring a 'read-before-write' sequence to avoid data loss, and mandating that multi-file changes be presented to the user for confirmation before execution.

mindos-zh