How to Use the Pre-Send Check Before Responses in i-have-adhd

The pre-send check in i-have-adhd automatically intercepts every model response before delivery, validating it against ten ADHD-friendly communication rules defined in README.md and enforced through the pipeline configured in .cursor/skills/i-have-adhd/SKILL.md.

The ayghri/i-have-adhd repository provides a Cursor skill designed to enforce strict formatting standards on AI-generated responses. The pre-send check serves as the quality gate, ensuring outputs lead with actionable steps, use numbered lists, and suppress tangents before reaching the user.

Understanding the Pre-Send Validation Pipeline

The pre-send check operates as a response hook implemented in .cursor/skills/i-have-adhd/SKILL.md. When the model generates a completion, the hook receives the raw text and validates it against ten specific rules—including leading with the next action, numbering steps, and suppressing tangents—before allowing transmission to the user interface.

If validation fails, the check blocks the non-compliant response and triggers a rewrite request, looping until the output satisfies all constraints. This mechanism guarantees that every delivered message adheres to the concise, action-first standards defined in the project's README.md.

Activating the Pre-Send Check

You can enable the validation pipeline through two primary methods: implicit activation via plugin installation or explicit configuration in agent-specific files.

Implicit Activation via Plugin Installation

When you install the skill as a plugin through supported agents like Claude or Codex, the pre-send hook attaches automatically to the response pipeline. The plugin.json file at the repository root declares the skill metadata, causing the agent to load .cursor/skills/i-have-adhd/SKILL.md and activate the check without additional configuration.

Explicit Configuration in Agent Files

For granular control, set the pre_send: true flag in your agent configuration files. The repository includes examples for multiple backends:

• skills/i-have-adhd/agents/openai.yaml – YAML configuration for OpenAI-compatible agents
• skills/i-have-adhd/agents/gemini.toml – TOML configuration for Gemini-based agents

Configuration Examples

Enable the check in your OpenAI agent by adding the skill and pre_send directive:

# skills/i-have-adhd/agents/openai.yaml

agents:
  openai:
    model: gpt-4
    skills:
      - i-have-adhd
    pre_send: true

For Gemini-based configurations, define the equivalent setting in TOML format:

# skills/i-have-adhd/agents/gemini.toml

[agent]
name = "gemini"
skills = ["i-have-adhd"]
pre_send = true

To verify activation via command-line tools, install the plugin and invoke the check explicitly:

# Install the skill via CLI

claude plugin install ayghri/i-have-adhd

# The check now runs automatically on every response

# Force immediate validation with:

/i-have-adhd

Summary

• The pre-send check is implemented in .cursor/skills/i-have-adhd/SKILL.md and validates responses against ten rules from README.md.
• Implicit activation occurs automatically when the skill is loaded via plugin.json.
• Explicit activation requires setting pre_send: true in agents/openai.yaml or agents/gemini.toml.
• Violations trigger automatic rewrite loops until the output complies with ADHD-friendly formatting standards.

Frequently Asked Questions

Where is the pre-send check logic implemented?

The validation logic resides in .cursor/skills/i-have-adhd/SKILL.md, which defines the rule set and hook mechanism that intercepts responses before delivery according to the repository source code.

Can I disable the pre-send check for specific responses?

Yes. Set pre_send: false in your agent configuration file, or temporarily disable the skill in your plugin manager to bypass validation for non-critical outputs.

Which specific rules does the check enforce?

The check validates against ten rules detailed in README.md, including leading with the next action, numbering sequential steps, suppressing tangential commentary, and ensuring concise, scannable formatting.

Is the check compatible with all AI agents?

The check works with any agent that supports the skill plugin architecture defined in plugin.json, including Claude, Codex, and OpenAI-compatible backends configured via the provided YAML and TOML examples.

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