The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read and build upon an existing plan.md file, following its contents as a strict roadmap for implementation.
Ingestion points: Reads plan.md from the project root directory during initialization.
Boundary markers: The instructions lack explicit markers or warnings to disregard potential malicious instructions or override attempts embedded within the processed plan file.
Capability inventory: The skill's functionality is primarily restricted to conversational reasoning and markdown file updates; it explicitly forbids code execution or modifications to other project files.
Sanitization: The skill does not specify any validation or sanitization of the data read from the plan file before it is incorporated into the agent's context.

plan-mode