The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from local files that may be influenced by external trade plans and incorporates this data into its own output, creating an indirect injection surface.
Ingestion points: The script size.py reads trade information from trader/plan.json and trader/prices.json. These files typically contain data generated by other skills or external market data.
Boundary markers: No specific delimiters or boundary markers are used to isolate potentially untrusted content within the JSON data structures during reading or writing operations.
Capability inventory: The skill possesses file-writing capabilities, specifically writing proposed bracket orders to trader/orders.json. This output is intended to be consumed by downstream skills such as trader-risk and trader-broker.
Sanitization: Input fields from the trade plan, such as symbol and setup, are interpolated directly into the why descriptive field in the trader/orders.json output. If these fields contain instruction-like text, they could potentially influence the behavior of agents or skills that process the order proposals later in the workflow.

trader-budget