| name | nd-taskboundary |
| description | Mark explicit cognitive task boundaries when topics shift or an interruption risks losing state. Save the active outcome, last verified fact, current action, blocker, and exact re-entry step before beginning the new context. |
Task Boundary — Clean Cognitive Transitions
What this addresses
Cognitive set persistence — a prior task's mental context bleeds into a new task. For many neurodivergent developers, switching tasks mid-session without an explicit transition leaves the old "cognitive set" (the assumptions, constraints, and vocabulary of the previous task) active. This contaminates the new task with irrelevant prior context.
Unlike ending the session or clearing history, a task boundary preserves what happened and explicitly marks where one cognitive set ends and another begins.
When to apply
Apply this skill when:
- The user's new message is substantively about a different topic than the active goal
- The user explicitly says something like "switching gears", "new thing", "different task", or "while I'm thinking about it"
- The topic shift would benefit from a clean mental context (not a /clear — you want to keep the history)
- A long session has just resolved one task and is about to start another
Signals of a topic shift (keyword-based heuristic)
- Less than 15% word overlap between the new request and the active goal statement
- New domain vocabulary that has not appeared in the last 5 turns
- A different software component, service, or codebase is suddenly the subject
Instructions
-
Detect the shift. Before the agent processes the new request, notice the topic has changed.
-
Name the transition (one line only, before responding to the new request):
"↩ Topic shift — moving from [prior task summary] to [new task]. Marking the boundary."
-
Save a restartable state before switching:
- active outcome
- last verified fact or artifact
- current action
- blocker or unresolved decision
- exact re-entry step
-
Insert the boundary marker into the conversation context:
[TASK BOUNDARY — new cognitive set begins here]
Previous task: [one sentence summary]
New task: [one sentence framing of the new request]
-
Begin the new task clean. Do not carry forward assumptions from the prior task unless explicitly told to. Treat the new task as if starting fresh — while still having access to the conversation history above the boundary.
-
On return: Resume from the saved re-entry step rather than reconstructing the conversation. Do not call the transition complete until the saved state can actually restart the work.
-
If uncertain: Ask once:
"Are you switching tasks, or is this related to [current goal]?"
Science basis
Set shifting (cognitive flexibility) is a core executive function that is commonly impaired in ADHD, autism, and other neurodivergent profiles. The difficulty is not with the new task itself but with deactivating the mental set from the old task.
Research on set-shifting impairment shows that an explicit environmental cue — a visible marker of transition — significantly reduces the interference from prior task context. The boundary acts as a cognitive anchor: "this is where that stopped and this started."
The boundary is NOT a /clear. The history is preserved. The marker is the signal that old context assumptions no longer apply.
Kolberg and Nadeau's launch points and externalized time, Dawson and Guare's flexibility supports, and Neff and NESTL's transition guidance support making context shifts visible and restartable.