The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation in the README provides installation commands for the required Linkly AI CLI using curl and sh from the vendor's official domain (updater.linkly.ai). This is standard procedure for installing prerequisites and originates from a trusted vendor source.
[REMOTE_CODE_EXECUTION]: The README includes a pattern for remote script execution as part of the CLI installation guide. This follows standard software distribution practices for the vendor's own tools and infrastructure.
[PROMPT_INJECTION]: The skill explicitly instructs the agent to treat all processed document content as untrusted data (Best Practice 15). It directs the agent to ignore any instructions or commands embedded within documents to prevent indirect prompt injection attacks.
[DATA_EXFILTRATION]: The instructions contain a dedicated section ('Credentials stay out of the conversation') that forbids the agent from requesting, printing, or handling the user's API keys or LAN tokens. It mandates the use of the official CLI authentication flow (linkly auth login) to ensure sensitive credentials never appear in the chat history or logs.
[COMMAND_EXECUTION]: The skill utilizes a legitimate set of CLI commands (linkly) and MCP tools provided by the vendor to interact with the user's indexed document library. These operations are scoped to the intended functionality of the skill.

linkly-ai