Purpose of the Pre-Send Check in the i-have-adhd Skill

The pre-send check is a post-processing filter that removes non-essential phrasing, hedging language, and conversational filler from responses to ensure only concrete, actionable information reaches readers with ADHD.

The ayghri/i-have-adhd repository implements a specialized content moderation layer called the pre-send check in the i-have-adhd skill. This step executes immediately before response emission to enforce formatting rules that accommodate limited working memory and attention constraints. By applying the guidelines defined in skills/i-have-adhd/SKILL.md, the skill transforms verbose or ambiguous outputs into concise, decision-ready instructions.

Why the Pre-Send Check Matters for ADHD-Friendly Communication

Readers with ADHD often experience reduced working memory capacity, making it difficult to retain information through preambles or process multiple ideas simultaneously. The pre-send check addresses this by ensuring the first line of every response presents the immediate next action, while the last line confirms the completed result. This structural predictability eliminates the cognitive load of parsing social niceties or hunting for the actual instruction buried in conversational text.

Six Filtering Rules Applied During the Pre-Send Check

1. Stripping Opening Announcements

The filter removes introductory sentences that merely announce upcoming actions, such as "Let me show you how..." or "Here is the information you requested." These phrases consume working memory without providing actionable value, violating the principle that the first line must be the next action.

2. Removing Closing Prompts

Closing sentences like "Anything else?" or "Let me know if you need help" are deleted because they distract from the task confirmation. The final line must instead state the concrete result of the action taken, ensuring the last line serves as a completed result marker.

3. Deleting Sidebar Distractions

Phrases beginning with "by the way" or tangential comments that split attention are excised immediately. Side notes create mental branching that can cause readers to lose focus on the primary task sequence, breaking the linear flow necessary for task completion.

4. Eliminating Hedging Adverbs

Words that introduce uncertainty—such as "perhaps," "might," or "possibly"—are stripped to reduce decision paralysis. The skill favors decisive language that clearly states what will happen or what action to take, removing ambiguity that can stall users with ADHD.

5. Replacing Idioms with Literal Actions

Figurative phrases like "circle back" or "get the ball rolling" are replaced with concrete, descriptive actions. Literal wording parses faster and requires less cognitive overhead than metaphorical interpretation, which is crucial when working memory is limited.

6. Enforcing First-and-Last-Line Verification

After pruning, the system verifies that the first line communicates exactly what to do next, while the last line confirms what just happened. This bookend validation guarantees that the response remains useful even if the reader only skims the extremes, matching the design principle that extremes carry the complete message.

Source Code Location and Implementation

The canonical definition of these rules resides in skills/i-have-adhd/SKILL.md under the Pre-send check section. A mirror of this configuration exists at .cursor/skills/i-have-adhd/SKILL.md for the Cursor runtime environment, ensuring consistent behavior across different editor implementations. Both files define the post-processing pipeline that executes immediately before the skill emits its final response to the user.

Before and After: A Practical Example

Consider how the filter processes a typical verbose response into ADHD-optimized output.

Before the pre-send check:

Let me show you how to add a new column.

1. Open `schema.sql`.
2. Add the column definition.
3. Run the migration.

That's it! Anything else you'd like to do?

After the pre-send check:

Add a new column to `schema.sql`.

1. Open `schema.sql`.
2. Add the column definition.
3. Run the migration.

Column added successfully.

The post-processing removed the introductory announcement "Let me show you...", the closing engagement question "Anything else...", and any unnecessary hedging, leaving only the actionable command and a clear confirmation of completion.

Summary

• The pre-send check in the i-have-adhd skill acts as a final content filter immediately before response emission.
• It removes conversational bookends, hedging adverbs, figurative idioms, and "by the way" sidebars that split attention.
• Verification ensures first lines state the next action and last lines confirm the completed result for skimmability.
• Rules are canonically defined in skills/i-have-adhd/SKILL.md with a runtime mirror at .cursor/skills/i-have-adhd/SKILL.md.
• Output prioritizes concrete, literal language that minimizes cognitive load and decision paralysis for readers with ADHD.

Frequently Asked Questions

What triggers the pre-send check in the i-have-adhd skill?

The pre-send check triggers automatically as the final post-processing step immediately before the skill emits its response. It acts as a gatekeeper that evaluates and modifies generated content against the formatting rules defined in the SKILL.md file to ensure ADHD-friendly output.

Where is the pre-send check defined?

The filtering rules are canonically defined in skills/i-have-adhd/SKILL.md under the Pre-send check section. An operational copy exists at .cursor/skills/i-have-adhd/SKILL.md to support the Cursor IDE runtime, ensuring consistent ADHD-friendly formatting across different environments.

Why does the skill remove hedging words like "perhaps"?

Hedging adverbs introduce uncertainty that creates decision paralysis for readers with ADHD. By removing words such as "might," "possibly," or "perhaps," the skill delivers definitive instructions that reduce cognitive load and support immediate action without ambiguity.

How does the verification step ensure ADHD-friendly output?

After content pruning, the verification step checks that the response's first line specifies the immediate next action and the last line confirms the completed result. This bookend validation guarantees that even readers who skim only the beginning and end of a message receive complete, actionable information.

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