The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to fetch and process data from multiple external financial news and analytics websites (e.g., macrotrends.net, stockanalysis.com, aastocks.com). This processing of external, untrusted content creates a surface for indirect prompt injection attacks.
Ingestion points: External URLs listed in the data source tables for US, HK, A, and Taiwan stocks (SKILL.md).
Boundary markers: The skill mandates a specific output format and requires every key data point to be verified against two independent sources, which acts as a verification layer.
Capability inventory: The skill can execute local scripts in the tools/ directory and use common shell commands.
Sanitization: The skill relies on cross-referencing logic and user-side validation of the data cutoff date.
[COMMAND_EXECUTION]: The skill instructs the agent to run local Python tools (tools/twstock_data.py, tools/financial_rigor.py) and the date command to establish context and process financial metrics. These commands are integral to the skill's primary purpose and use project-specific relative paths.

financial-data