The Agent Skills Directory

[SAFE]: The skill implements a "capture-only" mode that explicitly instructs the agent to treat all user statements as data (todos) rather than instructions to execute. This is a recommended pattern for preventing prompt injection when handling untrusted user input.
[SAFE]: File system operations are scoped to the .claude/ directory for managing task lists and documentation. No access to sensitive system files, environment variables, or credentials was observed.
[SAFE]: No network operations, external dependencies, or remote code execution patterns are present in the skill instructions.

record-todos