The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external artifacts like logs, HAR files, and trace captures. While these are untrusted data sources, the skill includes explicit instructions for the agent to redact secrets and authentication headers before processing or displaying the content.
[DYNAMIC_EXECUTION]: The methodology involves generating and running diagnostic scripts, test suites, and automated bisection harnesses (e.g., git bisect). These operations are standard debugging practices and are performed within the agent's execution environment as part of the intended workflow.
[CREDENTIALS_UNSAFE]: The instructions specifically address the risk of credential exposure by mandating a redaction policy. It requires replacing secrets with a redacted placeholder and encourages building loops against environment variables to keep credentials out of the captured artifacts.

diagnosing-bugs