The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration were detected. All activities are confined to local repository analysis.- [COMMAND_EXECUTION]: The skill instructs the agent to use standard local commands such as git and ripgrep for repository inspection. These commands are necessary for the skill's primary purpose and do not include dangerous flags or network operations.- [DATA_EXPOSURE]: No patterns of data exfiltration or hardcoded credentials were found. Repository metadata and source code are accessed locally for the purpose of issue investigation, and there are no commands to send this data externally.- [INDIRECT_PROMPT_INJECTION]: As the skill reads content from the repository being investigated, there is a theoretical surface for indirect prompt injection if those files contain malicious instructions. The workflow focuses on manual inspection and reporting, which significantly reduces the risk of automated obedience to embedded instructions.

investigate-issue