The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow for reading codebase content to generate implementation plans, creating a potential surface for indirect prompt injection attacks where malicious instructions in the source code could influence the agent's planning output.
Ingestion points: The skill instructions specify using read_file and search_files to understand the project structure and key files in the workspace.
Boundary markers: There are no specific instructions or delimiters provided to ensure the agent ignores instructions or logic embedded within the files it reads.
Capability inventory: The skill utilizes write_file to save plans and references the subagent-driven-development skill for executing tasks, which involves the delegate_task capability.
Sanitization: The skill does not define methods for sanitizing, escaping, or validating the content extracted from the workspace before it is interpolated into the generated markdown plan.

plan