The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests structured financial data from the PandaData API, which presents a surface for indirect prompt injection. However, the data is processed as structured DataFrames and is not interpreted as executable instructions.
Ingestion points: Data fetching methods in panda_data.py and panda_future_tick.py.
Boundary markers: The skill does not implement specific delimiters to isolate external data from the prompt context.
Capability inventory: The skill can write files to disk and spawn worker processes for parallel data processing.
Sanitization: Symbol conversion and date validation are performed, but the raw data values from the API are not sanitized.
[EXTERNAL_DOWNLOADS]: The skill requires the panda_data library, which is a standard external dependency for this service integration. Users are instructed to install it from official package registries.
[COMMAND_EXECUTION]: The skill uses ProcessPoolExecutor to parallelize data downloads by spawning multiple Python processes. This is a standard approach for improving performance in data-heavy tasks.
[CREDENTIALS_UNSAFE]: Access to the PandaData API is managed through environment variables or local .env files. The skill correctly prioritizes existing environment variables and provides a safe mechanism for loading configuration.
[DYNAMIC_EXECUTION]: The skill uses deferred imports to load the service SDK at runtime. This is a common performance optimization and does not involve the execution of dynamically generated or untrusted code.

ingest