The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/verify_earnings.py executes a local helper script (mx_data.py) via subprocess.run to retrieve quarterly financial data.
Evidence: The function run_mx_data constructs a command list using sys.executable and a relative path to the mx-data tool, then executes it without a shell.
Note: This is a standard implementation for tool interaction within the agent environment and is protected against shell injection.
[EXTERNAL_DOWNLOADS]: The skill specifies a dependency on the openpyxl Python library in its configuration.
Evidence: The YAML frontmatter in SKILL.md lists openpyxl under the install section.
Note: This is a widely used and well-known package for processing Excel data files.
[PROMPT_INJECTION]: The skill processes financial growth metrics retrieved from external financial platforms (such as Eastmoney) to inform the agent's recommendation workflow.
Ingestion points: Quarterly earnings growth data fetched by the mx-data tool in scripts/verify_earnings.py.
Boundary markers: Data is integrated into the final output using a structured template in SKILL.md.
Capability inventory: Subprocess execution via scripts/verify_earnings.py.
Sanitization: The script extracts numeric percentage values from the data using regular expressions before applying comparison logic.
Note: As the data originates from well-known and reputable financial services, the risk of indirect prompt injection is minimal.

@3182/earnings-surprise-detector