todo-guardian

To enforce actionable TODOs, you need a validation mechanism that detects deferral language in AI output and blocks completion claims until each deferred item is justified. This captures loose ends automatically during workflow generation.

Deferral language detection scans AI output for phrases indicating delayed action, such as explicit mentions of fixing items later. It records each occurrence as a todo item, performing deduplication to avoid duplicate entries in your workflow validation process.

You capture deferred tasks by activating a validation skill that monitors AI output generation in real-time. It identifies deferral phrases, requires a mandatory note for each item, and uses add commands to record them.

Yes, you can block progress at milestones like commits, session ends, and completion claims. The validation tool audits captured items at these checkpoints and blocks progress if any deferred task lacks proper justification.

No specific dependencies or components are required to validate deferrals in your workflow. The enforcement mechanism operates independently using a dedicated shell tool with add and check commands to record and audit items.

The best way to stop untracked deferrals is to enforce capture during AI output generation. By requiring justification for each deferred task and auditing at milestones, you prevent the AI from making completion claims with unresolved loose ends.