Restating State in i-have-adhd to Manage Working Memory: Rule 5 Explained

The i-have-adhd skill mandates that assistants restate the current workflow state at the end of every response to compensate for the working memory limitations common in users with ADHD.

The open-source repository ayghri/i-have-adhd provides a specialized skill definition designed to reduce cognitive load during AI-assisted workflows. Restating state in i-have-adhd is governed by Rule 5, which explicitly requires assistants to recap progress and context at the end of each turn, ensuring users never lose track of their position in multi-step tasks. This pattern addresses the neurological reality that people with ADHD cannot reliably hold contextual information between messages without explicit reinforcement.

The "Restate State Every Turn" Rule in SKILL.md

The definitive specification for restating state in i-have-adhd resides in skills/i-have-adhd/SKILL.md at lines 73-80. This file contains the complete set of 10 interaction rules, with Rule 5 specifically dedicated to working memory support.

Rule 5 Specification

According to the source code, Rule 5 states:

"The reader cannot hold 'we are on step 3 of 5' between messages. Restate it."

This directive requires the assistant to explicitly mention the current position in any workflow (e.g., "Step 2 of 4 done: schema updated") before presenting the next action. The rule recognizes that information vanishing from the screen effectively disappears from working memory for users with ADHD, making scroll-back navigation cognitively expensive.

Working Memory Constraints

People with ADHD possess limited working-memory capacity, meaning they cannot rely on information that is no longer visible. When an assistant fails to restate state, the user must either remember previous context (high cognitive load) or scroll back through the conversation (context switching penalty). The "restate state every turn" rule eliminates this burden by making every response self-contained.

Implementation Examples

When implementing restating state in i-have-adhd, assistants must format responses to include both the completed progress and the upcoming action. Here are concrete examples contrasting compliant and non-compliant outputs.

Correct State Restatement

// Compliant output following Rule 5
Step 2 of 4 done: added `last_login` column to the users table.
Next: back-fill the new column with existing data. Run:
    npm run migrate -- step=backfill

This example explicitly states the current step (2 of 4), confirms the completed action, and then transitions to the next instruction.

Incorrect Output (Missing Restatement)

// Non-compliant output - violates Rule 5
Added the `last_login` column.

Run:
    npm run migrate -- step=backfill

This version forces the user to remember that this is step 2 of 4 and what was accomplished in step 1, creating a "lost-in-the-middle" error risk.

Helper Function for Implementation

Developers building tools that generate these responses can use utility functions to ensure consistent formatting:

def restate_state(step, total, description):
    return f"Step {step} of {total} done: {description}."

# Usage example

msg = restate_state(3, 5, "schema updated")
print(msg)   # → "Step 3 of 5 done: schema updated."

Integration with Complementary Rules

Restating state in i-have-adhd does not operate in isolation. It works synergistically with other rules defined in the skill files to create a tight, action-oriented interaction loop.

Lead-with-Next-Action Principle

While Rule 5 handles backward-looking context ("what we just did"), the lead-with-next-action rule handles forward-looking guidance ("what to do now"). Together, they bracket the user's current position with clear before-and-after context, minimizing ambiguity about immediate priorities.

Number-Multi-Step-Tasks Rule

The number-multi-step-tasks rule ensures that workflows are divided into countable segments (e.g., 1 of 5, 2 of 5), providing the numerical framework that makes restating state meaningful. Without numbered steps, phrases like "we are on step 3" lack referential clarity, rendering the restatement less effective.

Summary

• Restating state in i-have-adhd is defined in skills/i-have-adhd/SKILL.md lines 73-80 as Rule 5.
• The rule requires explicit progress recaps at the end of every assistant response to support ADHD working memory limitations.
• Compliant responses include the step number, total steps, and description of completed work (e.g., "Step 2 of 4 done").
• This pattern prevents "lost-in-the-middle" errors by ensuring no response assumes the user remembers prior context.
• Rule 5 integrates with lead-with-next-action and number-multi-step-tasks to create self-contained, actionable messages that reduce cognitive load.

Frequently Asked Questions

Where is the rule for restating state defined in i-have-adhd?

The rule for restating state is formally defined in skills/i-have-adhd/SKILL.md at lines 73-80. This file contains Rule 5, titled "Restate state every turn," which explicitly requires assistants to recap the current workflow position at the end of each response. Developers can also reference README.md for a summary of all 10 rules and AGENTS.md for guidance on modifying the skill behavior.

Why does i-have-adhd require state restatement on every single turn?

The requirement exists because users with ADHD have limited working-memory capacity that does not retain information once it leaves the screen. According to the skill definition, "The reader cannot hold 'we are on step 3 of 5' between messages." Without explicit restatement, users must either scroll back (disrupting flow) or rely on memory (error-prone), both of which increase cognitive load and the risk of context loss.

How does restating state differ from simply numbering steps?

Numbering steps (defined in the number-multi-step-tasks rule) divides work into countable segments, while restating state (Rule 5) requires the assistant to verbally confirm those numbers at the end of each response. A workflow might internally be step 3 of 5, but unless the assistant explicitly states "Step 3 of 5 done," the user lacks the external memory cue needed to maintain context. The two rules work together: numbering provides structure, while restatement provides continuous reinforcement.

What happens if an assistant violates the state restatement rule?

Violations create "lost-in-the-middle" errors where the user becomes uncertain about what has been accomplished and what remains. For example, if an assistant simply says "Added the column" without stating "Step 2 of 4 done," the user must either remember the step count or scroll back to verify progress. This friction contradicts the skill's core mission of reducing cognitive load for ADHD users and can lead to task abandonment or errors in multi-step workflows.

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