The Agent Skills Directory

[SAFE]: The skill is a specification for building a tool and does not bundle any executable source code or binaries.
[EXTERNAL_DOWNLOADS]: The instructions require the agent to install standard Python packages such as grvt-pysdk, httpx, and requests from official package registries. These dependencies are necessary for the bot's documented functionality.
[COMMAND_EXECUTION]: The skill utilizes local shell commands for project scaffolding and management. It includes a specific curl command to interact with the Telegram API for initial configuration discovery, which is a standard procedure for bot setup.
[CREDENTIALS_UNSAFE]: The skill implements secure credential management by instructing the agent to store API keys in a local .env file and specifically advising against echoing these secrets in the chat interface.
[DATA_EXFILTRATION]: The bot is designed to send trading orders to the GRVT exchange and notifications to Telegram. These network operations are directed to the intended services and are fundamental to the bot's core purpose.
[PROMPT_INJECTION]: The architecture includes an indirect prompt injection surface where external data is ingested to trigger actions.
Ingestion points: The receiver process monitors Server-Sent Events (SSE) from the vendor's API stream (api.elfa.ai/v2/auto/queries/{id}/stream).
Boundary markers: The implementation includes logic to verify that the queryId in the incoming data frame matches the expected strategy ID.
Capability inventory: The bot can place atomic orders on the GRVT exchange and transmit alerts to Telegram.
Sanitization: The parser validates the JSON schema and checks specific field values before proceeding with order placement.

elfa-grvt-bot