The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides interfaces that retrieve untrusted, user-generated or third-party text data from financial information sources.\n
Ingestion points: Data enters the context via the irm_qa_sh, irm_qa_sz, news, and anns_d interfaces documented in the reference/ directory.\n
Boundary markers: Absent. There are no instructions in SKILL.md to wrap this external content in delimiters or warn the model to ignore embedded instructions.\n
Capability inventory: The skill uses the Bash tool to run Python scripts, providing a capability that could be misused if instructions were injected.\n
Sanitization: Absent. Content is retrieved and presented without filtering or escaping.\n- [PERSISTENCE]: The README.md suggests that users persist their Tushare Token by adding an export statement to their ~/.bashrc file, which is a documented pattern for modifying shell startup scripts.\n- [EXTERNAL_DOWNLOADS]: The skill requires external Python libraries (tushare, pandas, openpyxl) from standard package registries as part of its setup.\n- [COMMAND_EXECUTION]: The skill instructions include commands to verify the Python environment and install dependencies using Bash.

tushare-finance