The Agent Skills Directory

[COMMAND_EXECUTION]: The launcher script trader_memory_cli.py uses subprocess.call to execute internal sibling scripts (thesis_store.py, thesis_ingest.py, thesis_review.py) within the skill's environment. It can optionally route these calls through the uv package manager if available, ensuring dependencies are met. This execution is limited to known local scripts.
[EXTERNAL_DOWNLOADS]: The fmp_price_adapter.py component performs network requests to financialmodelingprep.com to retrieve historical price data for stocks. These requests are used to calculate performance metrics and require an API key supplied by the user via environment variables.
[INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface by ingesting data from external files produced by other agent skills.
Ingestion points: The thesis_ingest.py script reads JSON and CSV files provided via the --input or --bulk-csv command-line arguments.
Boundary markers: Structural boundaries are enforced through strict jsonschema validation defined in schemas/thesis.schema.json before any data is persisted.
Capability inventory: The skill has the capability to write to the local file system (state directory) and perform network GET requests.
Sanitization: Inputs are parsed and mapped to a fixed internal schema, and numeric fields undergo range and finiteness checks to prevent malformed data from affecting calculations.

trader-memory-core