The Agent Skills Directory

[SAFE]: The skill and its associated scripts and reference files were analyzed and found to contain no malicious code, obfuscation, or unauthorized data exfiltration logic.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and process untrusted repository data (source code, CLAUDE.md) using tools like Read, Grep, and Glob. However, the risk is mitigated by the skill's internal logic which requires evidence-based validation for all findings.\n- [COMMAND_EXECUTION]: The skill utilizes powerful tools such as Bash and Write to interact with the project environment. This access is consistent with the skill's purpose of debugging and testing, and is explicitly defined in the platform configuration.

systematic-debugging