The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user data (issue descriptions and file paths) and processes them to perform analysis and generate reports.
Ingestion points: User-supplied issue descriptions and file paths extracted in Step 1 of the workflow.
Boundary markers: The instructions do not define clear delimiters or use 'ignore embedded instructions' warnings when the agent processes user-provided issue text.
Capability inventory: The skill has read/write access to the local file system using tools like Read, Write, Grep, and Glob.
Sanitization: There is no mention of sanitizing or validating user-provided keywords or paths before they are used in tool operations.
[COMMAND_EXECUTION]: The skill uses the Grep tool to locate code blocks using keywords derived directly from user input, which could be manipulated to execute unintended searches or potentially influence tool execution depending on how the underlying platform handles shell arguments.
Evidence: '使用 Grep 搜索关键词找到相关代码' and 'Grep 工具：搜索函数定义... 搜索函数调用' in the execution steps of SKILL.md.

issue-trace