The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it is designed to ingest and analyze untrusted external data from public sources (e.g., A-share company announcements and financial news).
Ingestion points: Data is ingested through tools like stock_notice_report, stock_individual_info_em, and stock_zh_a_hist as described in references/methodology.md.
Boundary markers: Absent. The instructions do not specify any delimiters or safety prompts to prevent the agent from following malicious instructions hidden within the ingested financial data.
Capability inventory: The skill can execute local Python scripts and shell commands to process data, as defined in references/data-queries.md.
Sanitization: Absent. No evidence of input validation or sanitization for external data content is provided.
[COMMAND_EXECUTION]: The skill instructions involve the execution of shell commands and Python scripts located in sibling directories.
Evidence: references/data-queries.md instructs the agent to activate a virtual environment (source ../../.venv/bin/activate) and execute a runner script (python ../findata-toolkit-cn/scripts/views_runner.py). These scripts are part of the findata-toolkit-cn resource, which is associated with the author's workflow.
[EXTERNAL_DOWNLOADS]: The skill initiates the installation of external dependencies during its environment setup.
Evidence: In references/data-queries.md, the command python -m pip install -r ../findata-toolkit-cn/requirements.txt is used to install Python packages. While these dependencies are for the vendor's own toolkit, they are pulled from public registries at runtime.

investment-memo-generator