How i‑have‑ADHD Handles Lists in Responses: Rules and Implementation

The i‑have‑ADHD skill shapes every model reply using two declarative list rules: it mandates numbered lists for multi‑step tasks where each item describes a single bounded action, and it caps all lists at five items to reduce cognitive load.

The ayghri/i‑have‑adhd repository provides a conversational skill that optimizes AI responses for readers with ADHD. Rather than applying runtime validation, the system relies on static declarative rules stored in a skill definition file to govern how the model structures lists during text generation.

Declarative List Rules in SKILL.md

The core list‑handling logic resides in skills/i‑have‑adhd/SKILL.md. This file contains static rules that guide the language model's output format, specifically targeting cognitive load management through two primary constraints defined at specific line ranges.

Numbering Multi‑Step Tasks

Lines 42‑55 of skills/i‑have‑adhd/SKILL.md mandate that any response requiring multiple actions must use a numbered list. Each line must describe a single, bounded step rather than combining multiple operations. The rule explicitly directs the model to keep lists as short as possible by folding trivial steps into preceding items, preventing the fragmentation of simple workflows into unnecessarily granular tasks.

The Five‑Item Cap

To prevent overwhelming the reader, lines 103‑106 impose a strict ceiling: lists must never exceed five items. When content naturally requires more than five steps, the rule requires the model to split the list into logical segments—such as "do now" versus "later"—or reduce the content to a top‑five ranking. This constraint ensures that working memory demands remain manageable throughout the interaction.

Runtime Implementation in the Extension

The extensions/i‑have‑adhd.ts file activates these rules by loading the skill definition and injecting it into the conversation context. When ADHD mode is enabled, the extension calls loadRules() to read the entire contents of SKILL.md (lines 46‑64) and appends these instructions to the system prompt.

Notably, the implementation relies on declarative guidance rather than programmatic enforcement. The rules exist as text instructions within the skill file; the model follows them during generation, but there is no post‑processing validation that parses or truncates lists programmatically. This design delegates formatting responsibility to the language model while keeping the extension lightweight and stateless.

Practical Examples of List Formatting

The following examples demonstrate the application of these rules, showing violations, correct formatting, and the split‑list strategy for lengthy procedures.

Incorrect formatting violates both the single‑action and five‑item constraints:

1. Open the repository, locate the config file, edit the settings, run the build, test the output, verify the deployment, clean up temporary files.

Correct formatting uses single bounded actions and respects the length cap:

1. Open `src/config.ts`.
2. Change `debug` to `false`.
3. Run `npm run build`.
4. Execute `npm test`.
5. Deploy with `npm run deploy`.

When more than five actions are necessary, the model must segment the workflow:

**Do now**
1. Open `src/config.ts`.
2. Change `debug` to `false`.
3. Run `npm run build`.

**Later**
4. Execute `npm test`.
5. Deploy with `npm run deploy`.
6. Clean up temporary files.

Summary

• The list‑handling rules reside in skills/i‑have‑adhd/SKILL.md as declarative text instructions.
• Rule one (lines 42‑55) requires numbered lists for multi‑step tasks, with one bounded action per line.
• Rule two (lines 103‑106) caps all lists at five items, requiring segmentation or prioritization for longer content.
• extensions/i‑have‑adhd.ts loads these rules via loadRules() and injects them into the context at lines 46‑64.
• The system uses prompt‑based guidance rather than runtime validation to shape list formatting.

Frequently Asked Questions

Where are the list rules defined in i‑have‑adhd?

The list rules are defined in skills/i‑have‑adhd/SKILL.md. The numbering rule appears at lines 42‑55, while the five‑item cap is specified at lines 103‑106. These rules are written as declarative instructions for the language model rather than as executable code.

Does i‑have‑adhd enforce list length programmatically?

No. According to the source code in extensions/i‑have‑adhd.ts, the extension loads the rules via loadRules() and injects them into the conversation context. The model follows these guidelines during text generation, but there is no runtime validation or truncation logic that enforces the five‑item limit after the text is produced.

How does i‑have‑adhd handle lists with more than five items?

When content exceeds five steps, the skill directs the model to split the list into logical groups—such as "do now" and "later"—or to prioritize the top five most critical items. This approach maintains the cognitive accessibility constraints while preserving all necessary information across multiple segmented lists.

What happens if a step is too trivial to stand alone?

The SKILL.md file instructs the model to fold trivial steps into preceding items rather than expanding the list unnecessarily. This ensures that the numbered list remains concise and that each item represents a meaningful, bounded action rather than fragmenting simple workflows into overly granular steps.

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