The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the tgcloud-mcp package globally and use the @tgcloud/bot scaffold. These packages originate from the platform's infrastructure and are necessary for the skill's primary function of managing Telegram serverless bots.
[COMMAND_EXECUTION]: The skill utilizes several CLI commands (e.g., push, migrate, run_handler, webhook_sync) to manage the project lifecycle. It incorporates critical guardrails by requiring explicit user confirmation before executing potentially destructive operations such as reset, push force, and non-dry-run migrations.
[INDIRECT_PROMPT_INJECTION]: The skill processes Telegram updates that act as ingestion points for untrusted external data, which could potentially influence agent behavior during testing and debugging cycles.
Ingestion points: The run_handler tool accepts arbitrary input in the args parameter (e.g., text, chat objects) meant to simulate Telegram bot updates.
Boundary markers: The instructions do not specify the use of delimiters or warnings to the agent to ignore instructions embedded within the simulated update data.
Capability inventory: The skill environment provides capabilities for network requests (fetch), database queries (db), and deployment actions (push, migrate).
Sanitization: There are no explicit instructions for the agent to sanitize or validate the external content before processing it.
[CREDENTIALS_SAFE]: The skill documentation includes best practices for secret management, specifically instructing that CLI access tokens should only be used with the login_bot tool and never printed, committed to version control, or stored in plain text files. It also correctly advises against committing the .tgcloud/ directory, which contains sensitive credentials and cache.

tgcloud