The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface as it ingests untrusted data from external search results and financial news via the argo and eastmoney data sources.
Ingestion points: Data enters the agent context through invest-cli/scripts/sources/argo.py (search results) and invest-cli/scripts/sources/eastmoney.py (news/company data).
Boundary markers: Limited boundary markers or explicit instructions to ignore embedded prompts in processed data were identified in the instruction files.
Capability inventory: The skill can execute shell commands and spawn subprocesses via invest-cli/scripts/invest_cli.py and its adapters to call external tools like ttskill or node.
Sanitization: No explicit sanitization or filtering of external content was observed in the Python scripts before returning data to the agent context.
[DYNAMIC_EXECUTION]: The skill uses dynamic module loading to manage its various data source adapters.
Evidence: invest-cli/scripts/sources/route.py and invest-cli/scripts/cmd_intent.py use importlib.import_module and __import__ to load adapters defined in data-sources.yaml. This is a standard architectural pattern for this skill's aggregator functionality and relies on internal configuration.
[COMMAND_EXECUTION]: The skill invokes external command-line tools and scripts to fulfill its purpose.
Evidence: Several files in invest-cli/scripts/sources/ (e.g., ttskill.py, yingmi.py, wind.py, argo.py) use subprocess.run to call tools like ttskill, yingmi-skill-cli, node, and python3. These calls are used to interact with official financial data connectors or authorized sibling skills.

fund-investment-guide