The Agent Skills Directory

Potential for Indirect Prompt Injection: The skill's exploration phase (Step 1) involves reading and analyzing the existing codebase to identify patterns and architecture. This creates an interface where the agent might ingest instructions embedded in the project files. While the skill defines clear boundaries for planning and implementation, users should be aware that the agent's output could be influenced by content found within the files it processes.
Ingestion points: The agent reads project files and documentation during the "Explore" phase.
Boundary markers: The skill does not explicitly define delimiters for separating untrusted codebase content from the planning instructions.
Capability inventory: The skill is designed to write a PLAN.md file and modify project files during implementation.
Sanitization: The instructions do not specify sanitization or validation of the content discovered during exploration.
File System Operations: The skill utilizes the agent's capability to read from and write to the local file system, specifically for creating a PLAN.md file and performing implementation tasks. These operations are core to the skill's purpose and are governed by the agent's execution environment permissions.
Command Execution in Testing: The planning process includes defining verification steps and running test commands. Users should review these commands in the generated plan before they are executed to ensure they align with the project's security and testing policies.

plan-mode