The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is granted access to the filesystem through tools like Read, Grep, and Bash. This access is necessary for analyzing existing code and generating test suites. Users should ensure the execution environment is properly scoped to the project directory.
[INDIRECT_PROMPT_INJECTION]: The skill processes project source code and configuration files to frame the test target (SKILL.md, Step 1). It uses tools like Read, Glob, and Grep to ingest this data. While the instructions do not include explicit boundary markers or sanitization logic to handle potentially malicious content within the code (such as instructions hidden in comments), this is a common characteristic of code-analysis skills. The agent's capabilities include Write, Edit, and Bash (SKILL.md, allowed-tools), which are used to implement the generated test plans.

backend-testing