The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a simple echo command using the Bash tool in the preamble section to notify the user that the risk management tool has started. This is a benign use of system tools for status reporting.
[INDIRECT_PROMPT_INJECTION]: The skill collects data from users via the AskUserQuestion tool across multiple steps and incorporates this untrusted data into a markdown report generated with the Write tool.
Ingestion points: User responses to questions about risk types, assessment methods, strategies, and monitoring frequencies in SKILL.md (Steps 1-4).
Boundary markers: There are no explicit delimiters or specific instructions to the agent to treat user-provided data as potentially malicious or to ignore embedded instructions within that data.
Capability inventory: The skill possesses Write capabilities for file creation and Bash capabilities for command line interaction.
Sanitization: The skill lacks logic to sanitize, validate, or escape user input before writing it to the final document at docs/04-风控管理/风险管控方案.md.

pm-risk