The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/run_stress_test.py uses a custom dynamic import mechanism to load analysis modules at runtime.
Evidence: The script utilizes a try_import function to conditionally load modules like src.core.risk.correlation and src.core.risk.shock_sensitivity based on their availability in the local environment.
Evidence: The script modifies the Python search path using sys.path.insert(0, ...) to include parent directories for importing local source files.
[INDIRECT_PROMPT_INJECTION]: The skill processes external financial data and user-provided scenarios, which are then interpreted by the AI agent to generate qualitative analysis.
Ingestion points: The skill ingests ticker symbols and scenario descriptions via command-line arguments in SKILL.md and fetches external data through yahoo_client in scripts/run_stress_test.py.
Boundary markers: The skill defines a 10-step structured output pipeline in SKILL.md to guide the agent's response format.
Capability inventory: The skill has access to the Bash tool and performs local file writes to save test history via src.data.history.save_stress_test.
Sanitization: Input symbols are split and stripped, and weights are converted to floats; however, external data from the stock API is processed directly by the agent for "LLM interpretation" and "LLM complement" steps.

stress-test