The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses standard instructional language to define the agent's persona as an A/B testing expert. There are no attempts to bypass safety filters, override system constraints, or extract internal prompts.
[DATA_EXFILTRATION]: No sensitive file paths or credentials were detected. The skill references .claude/product-marketing-context.md, which is a standard context file for the target platform. There are no network operations that could exfiltrate data.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute remote code. It includes a local script, sample_size_calculator.py, which is implemented using only the Python standard library and performs mathematical calculations locally.
[COMMAND_EXECUTION]: The skill instructs the agent to use the provided sample_size_calculator.py script. This is a local, controlled execution environment for a specific analytical task and does not allow for arbitrary command injection.
[OBFUSCATION]: Analysis of the skill's text and scripts revealed no base64 encoding, zero-width characters, homoglyphs, or other techniques used to hide malicious content.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from .claude/product-marketing-context.md to inform test design. This is a standard integration point for the platform and does not present an unusual attack surface, as the skill produces structured A/B testing artifacts rather than executing commands based on that data.

ab-test-setup