The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from local files to generate trading platform code, which creates a surface for indirect prompt injection attacks.
Ingestion points: The agent is instructed to read config.json and code/signal_engine.py (referenced in SKILL.md) to understand strategy logic.
Boundary markers: No specific delimiters or "ignore instructions" warnings are provided to wrap the ingested content before it is processed by the model.
Capability inventory: The skill utilizes read_file to access strategy logic and write_file to save generated artifacts to the local file system.
Sanitization: There is no evidence of sanitization or validation of the input logic or parameters before they are used to generate code.

pine-script