The Agent Skills Directory

[Potential Indirect Prompt Injection]: The skill's workflow involves gathering data from external and potentially untrusted sources, such as public benchmarks and team communications, which creates a surface for indirect prompt injection.
Ingestion points: Data is gathered from ~~company_docs, ~~team_communication, and various public sources (SKILL.md).
Boundary markers: The instructions do not specify the use of clear delimiters to separate external content from the agent's primary instructions.
Capability inventory: The skill has the ability to generate persistent artifacts via $build-report and $Spreadsheets, and can execute code using $jupyter-notebooks.
Sanitization: There are no explicit requirements for sanitizing or filtering data retrieved from external sources.
[Dynamic Code Execution]: The skill utilizes tools to execute code for data processing and modeling.
Execution pattern: The instructions call for the use of $jupyter-notebooks to handle source harmonization, calculations, and sensitivity analysis.
Context: This behavior is intended for the skill's analytical purposes but involves the runtime generation and execution of logic that may be influenced by input data.

market-sizing