The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script located at tools/financial_rigor.py with subcommands such as verify-valuation and three-scenario to perform financial arithmetic and valuation checks. This is a core part of its functionality for ensuring data accuracy.
[EXTERNAL_DOWNLOADS]: The skill uses a WebSearch agent to retrieve real-time market data, including stock prices, valuation multiples, and recent financial news from the internet.
[DATA_EXFILTRATION]: The skill reads from and writes to a local file named reports/portfolio-latest.md to maintain a persistent record of the user's portfolio and analysis results. It instructions the agent to update this file during the execution flow.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes untrusted data from both user input ($ARGUMENTS) and external web search results.
Ingestion points: Data enters the context through user-provided portfolio descriptions and external content retrieved via the WebSearch tool.
Boundary markers: None explicitly defined in the instructions; the agent relies on structured markdown tables for data processing.
Capability inventory: The agent can execute shell commands (running the local financial tools) and write to local files (the portfolio report).
Sanitization: No explicit sanitization or filtering logic is defined within the skill instructions for external content.

portfolio-review