The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted content from the local codebase being analyzed. If the codebase contains malicious instructions, the agent or its sub-agents could potentially be influenced during the design or issue creation phase.- Ingestion points: Reads local codebase files during the exploration phase (Step 1) using the Agent tool.- Boundary markers: No specific boundary markers or 'ignore' instructions are provided to delimit codebase data from agent instructions.- Capability inventory: Employs sub-agents for design tasks and uses the gh CLI to create GitHub issues.- Sanitization: No sanitization or validation of the codebase content is performed before processing.- [COMMAND_EXECUTION]: The skill utilizes the gh issue create command to autonomously create refactoring RFCs. The instructions explicitly direct the agent not to ask for user review before creating the issue, relying on the automated analysis performed in previous steps. This is a standard functional behavior for the intended use case.

improve-codebase