The Agent Skills Directory

[PROMPT_INJECTION]: The skill documents patterns for processing external input that could be used for indirect prompt injection attacks.
Ingestion points: External data from Telegram updates, such as message.text and query.data, are ingested in handlers and filters as shown in references/features-smart-plugins.md and references/features-filters-custom-filters.md.
Boundary markers: The provided code examples do not demonstrate the use of delimiters or explicit instructions to ignore embedded commands within the processed message content.
Capability inventory: Handlers have access to powerful bot capabilities across various scripts, including message.reply, message.forward, and client.invoke for raw API operations.
Sanitization: The examples do not include logic for sanitizing or validating external content before it is processed or echoed back to the user.
[REMOTE_CODE_EXECUTION]: The skill describes the "Smart Plugins" feature which utilizes dynamic loading of Python modules.
Dynamic Loading: The Client configuration in references/features-smart-plugins.md shows auto-discovery and loading of modules from a local plugins directory at runtime. While this is a standard library feature, it allows for the execution of code found in the specified path.
[EXTERNAL_DOWNLOADS]: The documentation recommends the use of external performance-enhancing packages.
Package Recommendations: references/best-practices-errors-sessions-performance.md suggests installing tgcrypto-pyrofork and uvloop to improve throughput. These are recognized as legitimate ecosystem components for this library.

pyrofork-best-practices