The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from web searches and local files, which could contain instructions designed to influence agent behavior.
Ingestion points: The skill ingests user input from $ARGUMENTS, financial data from WebSearch, and local data from reports/portfolio-latest.md.
Boundary markers: There are no explicit delimiters or specific instructions to ignore embedded commands within the processed data.
Capability inventory: The skill has the capability to write to local files (reports/portfolio-latest.md) and execute local Python scripts (tools/financial_rigor.py).
Sanitization: The instructions do not specify any sanitization or validation of the content retrieved from external sources before it is processed by tools or written to files.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent to use the system shell to run the date command and execute local scripts located in the tools/ directory. While these are presented as internal project tools, they constitute an execution surface that relies on the integrity of the local repository and the safety of the input arguments passed to them.

portfolio-review