The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external web searches and user inputs to perform financial analysis and drive logic, creating a surface for indirect instructions to influence agent behavior.
Ingestion points: Web search results (via WebSearch) and user-provided arguments ($ARGUMENTS).
Boundary markers: Absent; there are no explicit delimiters or instructions to ignore embedded commands within the retrieved data.
Capability inventory: The agent can execute shell commands (python3 tools/financial_rigor.py), read from and write to workspace files (reports/{公司名}-thesis.md), and perform further web searches.
Sanitization: None; the instructions do not specify validation or sanitization of content retrieved from external financial sites.
[COMMAND_EXECUTION]: The skill utilizes local shell commands to establish current context and verify financial metrics.
Evidence: Instructions to run the date command and execute vendor-provided tools located in the tools/ directory (e.g., python3 tools/financial_rigor.py verify-valuation).

thesis-tracker