The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script scripts/smartmoney.py using the exec tool to process market data. This is the primary intended function of the skill to perform deterministic calculations that the LLM cannot do alone.
[EXTERNAL_DOWNLOADS]: The skill uses an internal MCPClient in scripts/mcp_client.py to communicate with mcp.prod.senpi.ai. This is a vendor-owned service ('senpi-ai') used for fetching legitimate market data and does not constitute an untrusted remote download.
[DATA_EXPOSURE]: The skill requires a SENPI_AUTH_TOKEN environment variable to access the Senpi API. The instructions correctly advise the user that this token is needed for the discovery_* tools. There is no evidence of hardcoded credentials or exfiltration of sensitive local files.
[INDIRECT_PROMPT_INJECTION]: The skill processes external market data (trader positions, PnL, and leaderboard events). While this is an ingestion surface for untrusted data, the skill uses a deterministic Python engine to filter and aggregate this data into a structured JSON schema before it reaches the agent, significantly reducing the risk of prompt injection from data sources. Boundary markers are recommended for the agent's narration phase.

senpi-smart-money