The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The README.md file contains installation instructions that fetch Python scripts and configuration files from the vendor's official GitHub repository (Senpi-ai/senpi-skills) using curl. These resources are necessary for the strategy's operation.
[PROMPT_INJECTION]: The skill is subject to an indirect prompt injection surface (Category 8) due to its processing of external strategy and position data.
Ingestion points: Data enters the system via fetch_top_strategies and fetch_strategy_positions in scripts/cuckoo-producer.py, which call external tools to retrieve leaderboard and position details.
Boundary markers: The decision prompt in runtime.yaml interpolates the ingested signal data without utilizing specific delimiters or instructions to ignore potential commands embedded within the data values.
Capability inventory: The agent has the ability to perform financial transactions (opening and closing trading positions) and access sensitive account state through platform tools.
Sanitization: The skill applies basic type casting and case normalization to ingested data but lacks comprehensive sanitization or validation for malicious prompt injection payloads.
[COMMAND_EXECUTION]: The installation process involves executing shell commands to configure the environment and start the producer daemon as a persistent background process using nohup and disown.
[REMOTE_CODE_EXECUTION]: The tests/test_signal.py script utilizes dynamic module loading via importlib.util to execute code from the local scripts/cuckoo-producer.py file for unit testing purposes.

cuckoo-strategy