The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use financial tools for opening, modifying, and closing trading positions. These tools (e.g., strategy_create, edit_position) are standard for the platform's trading functionality and are used within the scope of the user's intent.
[PROMPT_INJECTION]: The instructions do not contain patterns typical of prompt injection, such as attempts to override safety filters, disregard previous instructions, or impersonate an unrestricted persona. The focus is strictly on executing trades and providing monitoring guidance.
[DATA_EXFILTRATION]: No evidence of data exfiltration was found. The skill does not attempt to access sensitive system files (e.g., SSH keys, AWS credentials) or perform unauthorized network requests. It explicitly instructs the agent never to send USDC to external addresses via tools.
[REMOTE_CODE_EXECUTION]: The skill does not include any patterns for downloading, installing, or executing remote scripts or packages. It relies on internal platform tools and referenced skills for its operations.
[CREDENTIALS_UNSAFE]: No hardcoded secrets, API keys, or private tokens were found in the skill metadata or body. The instructions correctly direct the use of environment-based or platform-managed configurations.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests external data (trader track records and position books), it provides specific instructions for validating this data (e.g., checking instrument casing, slippage gates, and win-rate anomalies). It treats external data as information to be vetted rather than instructions to be executed.

senpi-trade