The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data fetched from external financial APIs which is subsequently analyzed by an LLM. This represents a standard analytical surface where data is interpolated into a prompt.
Ingestion points: Market data is retrieved in core/data_fetcher.py and core/flow_fetcher.py.
Boundary markers: Present. Instructions in llm/analyst.py use structured markdown headers and delimiters to separate data from analysis instructions.
Capability inventory: The skill has capabilities to write reports to the local file system (core/reporter.py), make network requests to financial data and AI providers (core/data_fetcher.py, core/flow_fetcher.py), and execute a bundled validation script (run.py).
Sanitization: The skill primarily processes numeric and categorical data; text fields like stock names are processed as part of the analysis workflow.
[COMMAND_EXECUTION]: The entry point run.py uses subprocess.run() to execute a local Python script (scripts/validate_screener.py) for output validation. This is a benign use of subprocess execution for internal tooling that does not accept unvalidated user input.
[DATA_EXPOSURE]: The skill follows standard security practices by using environment variables for sensitive API credentials. No hardcoded secrets or sensitive path exposures were found.
[EXTERNAL_DOWNLOADS]: All network operations target well-known financial services (East Money) or established technology providers (Pandadata, Anthropic/DeepSeek). No downloads from untrusted sources were detected.

post-market-screener