The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing user queries and a 'business context' block stored in session memory to determine which tools to execute. This involves ingesting potentially untrusted data, such as customer complaints or external communications.
Ingestion points: User input queries and the ## Контекст бизнеса block in session memory (SKILL.md).
Boundary markers: The instructions do not specify any delimiters or safety markers for the interpolated context or user queries.
Capability inventory: The skill is designed to trigger numerous slash commands and external tools, including those with network access or file-writing capabilities (e.g., /run-campaign, /handle-complaint).
Sanitization: The skill relies on a user confirmation step ('Начинаем?') before executing commands, but lacks explicit instructions for the AI to sanitize or escape the data before passing it to the downstream tools.
[EXTERNAL_DOWNLOADS]: The instructions suggest the installation of supplementary plugins (humanizer-ru@small-business-ru and marketplaces-mcp-ru@small-business-ru) via the /plugin install command. These appear to be intended components of the localized business suite provided by the vendor.

smb-router