Why Rule 5: Restate State Every Turn Is Critical for ADHD-Friendly Interactions

Rule 5: Restate state every turn is essential because it compensates for limited working memory in ADHD users, provides explicit progress tracking, and ensures system resilience by embedding complete context in every interaction.

The ayghri/i-have-adhd repository defines a structured conversational skill designed specifically for users with ADHD. Rule 5—restating the current state on every turn—serves as a critical safeguard against context loss and cognitive overwhelm. This principle appears in README.md and is extensively documented in skills/i-have-adhd/SKILL.md.

What Rule 5: Restate State Every Turn Means

Rule 5 requires that every response in a multi-step interaction explicitly repeats the current progress position and overall context. Rather than assuming the user remembers previous steps, the assistant must state "Step X of Y" along with the current action and next steps. According to the specification in skills/i-have-adhd/SKILL.md, this transforms abstract sequences into concrete, trackable checkpoints.

Why This Rule Is Essential for ADHD Users

Memory-Size Limitations and Working Memory

People with ADHD typically exhibit a very small working-memory buffer. When a conversational thread progresses, anything not explicitly repeated is quickly forgotten. By restating the current step and overall progress on each turn, the skill refreshes the reader’s mental model, preventing loss of context. As noted in README.md at line 68, this explicit repetition directly addresses the cognitive characteristics of the target users.

Progress Visibility and Dopamine Motivation

Explicitly stating "Step X of Y" turns an abstract sequence into a concrete, trackable checklist. This makes the remaining workload evident, which is essential for maintaining motivation through dopamine regulation and for avoiding "analysis paralysis." The SKILL.md file (lines 75-81) emphasizes that visibility into progress prevents users from becoming overwhelmed by ambiguous next steps.

Error Resilience and Crash Recovery

If the assistant or downstream tool crashes, or if the user switches topics, the most recent turn still contains the complete state. The system can resume without guessing what step was last completed, eliminating ambiguous recovery paths. This self-contained state management ensures that no critical context is lost between interactions.

Harness Orchestration and State Synchronization

The skill is often invoked from a harness that expects a single active step at a time. By emitting the state on every turn, the harness’s checklist can stay synchronized automatically without needing extra state-tracking code elsewhere. As implemented in skills/i-have-adhd/SKILL.md (line 82), this consistent orchestration simplifies the integration between the skill and its execution environment.

How to Implement Restate State Every Turn

The skills/i-have-adhd/SKILL.md file (lines 33-64) outlines the action-first philosophy that this rule enforces. Below are practical implementations:

When sending plain text responses, always prefix with the step counter:

Step 2 of 4: Update the data schema.  
Next: run `npm run migrate` and verify the new column appears.

For tool consumptions that expect structured output, include explicit step metadata:

{
  "step": 3,
  "totalSteps": 5,
  "message": "Back‑fill the new column with default values.",
  "nextAction": "Execute `node scripts/backfill.js`"
}

When integrating with a harness that maintains a checklist, consume the state fields to update progress tracking:

// pseudo-code used by the harness
if (response.step && response.totalSteps) {
  checklist.update({
    current: response.step,
    total: response.totalSteps,
    description: response.message
  });
}

In each case, the response always repeats the current step and total steps, satisfying Rule 5.

Summary

• Rule 5: Restate state every turn compensates for ADHD-related working memory limitations by explicitly repeating context on every interaction.
• Progress visibility (Step X of Y) provides concrete checkpoints that reduce anxiety and maintain dopamine-driven motivation.
• Error resilience ensures that crashes or interruptions never result in lost context, as the most recent message contains all necessary state information.
• Harness synchronization becomes automatic when each turn emits complete state data, eliminating the need for complex external tracking logic.
• Implementation requires explicit step counters in either plain text or structured JSON formats, as defined in skills/i-have-adhd/SKILL.md.

Frequently Asked Questions

What happens if I skip Rule 5 and assume the user remembers the previous step?

Users with ADHD will likely lose track of the conversation context due to working memory limitations. According to the design principles in README.md, this leads to confusion and abandonment, as the cognitive load of reconstructing the sequence outweighs the task itself.

Can I implement Rule 5 with just a simple progress bar instead of text?

While visual indicators help, the rule specifically requires explicit verbal or textual restatement of the current step and total steps. A progress bar alone doesn't provide the concrete "Step X of Y" cognitive anchor that prevents analysis paralysis, as detailed in skills/i-have-adhd/SKILL.md (lines 75-81).

How does Rule 5 help when the system crashes mid-conversation?

Because every turn contains the complete state (current step, total steps, and next action), the system can resume immediately from the last message without maintaining external session state. This self-contained approach eliminates ambiguous recovery scenarios where the system must guess what the user last completed.

Is Rule 5 only relevant for users with ADHD?

While designed specifically for ADHD cognitive patterns, restating state every turn benefits all users by reducing cognitive load and clarifying progress. However, for the i-have-adhd skill, this rule is mandatory rather than optional, as it forms a core accessibility requirement for the target audience.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}