The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions define a specific persona and analytical framework for investment decision-making. There are no directives to override system safety guidelines, bypass ethical constraints, or extract internal system prompts. The instruction to 'proactively trigger' the persona for investment topics is a standard functional requirement for specialized agents.
[DATA_EXFILTRATION]: No network operations, such as curl or wget, were detected. The skill does not access sensitive local file paths (e.g., .ssh, .aws, .env). It explicitly restricts its file access to a local '/references/' directory using a provided read tool.
[COMMAND_EXECUTION]: The skill does not utilize shell commands, subprocesses, or binary execution. It operates entirely as an information retrieval and reasoning system based on the provided Markdown reference files.
[EXTERNAL_DOWNLOADS]: The skill is self-contained. It does not attempt to download external scripts, packages, or configurations from remote servers. All 'reading' is directed towards the included reference documentation.
[INDIRECT_PROMPT_INJECTION]: The skill processes external financial data and reports. While this establishes an ingestion surface for potentially untrusted data, the skill is restricted to analytical output within a strict framework. No capability for executing arbitrary actions based on ingested data (like automated trades or network requests) is present.

buffett