The Agent Skills Directory

[SAFE]: The skill includes proactive security instructions for data redaction. It explicitly mandates that all secrets, credentials, and authentication headers in captured artifacts (logs, HAR files, etc.) must be replaced with a placeholder before being processed or displayed.
[DYNAMIC_EXECUTION]: The skill methodology involves creating diagnostic scripts, test harnesses, and CLI wrappers to reproduce bugs. This dynamic creation of scripts is the primary intended function of the debugging skill and occurs within the controlled environment provided by the agent platform.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data sources such as codebase documentation (ADRs, GLOSSARY.md), system logs, and network traces. While these ingestion points represent an attack surface for indirect prompt injection, the risk is inherent to the debugging task, and the instructions prioritize user-controlled environments and HITL (Human-in-the-loop) verification.

diagnosing-bugs