1. The Core Bottleneck: What Engineering Flaw Does It Break?

Current AI coding assistants suffer from a severe tendency to dump redundant information. After a developer inputs a bug triage or architecture refactoring request, the large language model invariably outputs warm emotional affirmations at the beginning, lengthy implementation analyses in the middle, and polite well-wishes at the end. This conversational pattern disrupts developer flow state and burns expensive context tokens on every interaction. The ayghri/i-have-adhd project cuts straight through this engineering nuisance by enforcing strict output specifications that force the AI to drop all pleasantries and push executable operational steps directly to the terminal screen.

💡 Core Architecture Insight: By encoding conversational state-machine output constraints into an uncompromising instruction set, the LLM's narrative reasoning layer is bypassed, compressing generation results into a pure action-instruction stream.

2. Core Architecture and Data Flow Analysis

i-have-adhd injects target AI coding environments using a lightweight Skill/Plugin modular architecture. The backend relies on no additional persistent server-side daemons, leveraging the strong adherence of large language models to system prompts or skill files to restructure the output topology during client-side parsing. When a developer triggers an interaction request, the data flow passes through the local gateway and the injected rule filter layer, directly truncating the model's reflective text and forcing the generation engine to output responses conforming to 10 hard rules.

[ Developer Prompt ] ---> [ Claude / CLI Gateway ] ---> [ i-have-adhd SKILL.md ]
                                                              │
                                                              ▼
                           [ Action-First Stream Engine ] <--- [ Rule Filter ]

The engineering trade-off is extreme: it completely sacrifices user-facing conversational warmth and human-like empathy, devoting all optimization resources to the absolute execution efficiency of code writing. By limiting list lengths, banning pleasantries and recaps, and forcing action-first initializations, interaction throughput density increases threefold.

3. Technology Selection and Hardcore Benchmarks

Evaluation Dimension This Solution (i-have-adhd) Traditional Implementation Typical Competitor Solution Production ROI
Interaction Latency Extremely low (zero proxy) High (waiting for verbose text) Medium (filtered middleware) 40% reduction in wait anxiety
Token Consumption Minimal (70% text reduction) Bloated (excessive filler) Average (filtered post-gen) Lower API billing overhead
Rule Customization Pure Markdown text editing Complex backend refactoring Closed SaaS dashboard Modification cost drops to seconds
Environment Dependency Claude CLI / plugin support Self-hosted proxy servers Commercial SaaS client Zero extra ops complexity

The metrics in the table demonstrate that traditional AI assistants generate heavy text payloads during critical debugging tasks. i-have-adhd sacrifices explanatory text so developers can locate exact target code lines and command strings within a single second.

4. Hands-on Geek Practice: Building the Minimum Viable Loop from Scratch

Execute the following commands in your local terminal to complete the global installation of the i-have-adhd skill plugin:

# Drop upstream or conflicting local copies first
claude plugin uninstall i-have-adhd
claude plugin marketplace remove i-have-adhd

# Add your custom fork marketplace source to the local environment
claude plugin marketplace add your-username/i-have-adhd

# Install the production-grade skill plugin
claude plugin install i-have-adhd@i-have-adhd

After installation, restart your coding assistant terminal and explicitly invoke the skill within any codebase directory:

/i-have-adhd check type boundary issues in verifyToken function within src/auth.ts

The expected structure of the assistant's output will omit all greetings and directly expose actionable directives:

Run npm install jsonwebtoken@latest, then edit src/auth.ts:42.

1. Open src/auth.ts
2. Replace verifyToken (lines 42–58) with the snippet below
3. Run npm test -- auth.spec.ts

Next: paste the first failing line if any test fails.

5. Production Gotchas and Avoidance Strategies

Deploying these hard rules into high-intensity daily engineering requires monitoring the degradation of model adherence to concise instructions. When multi-turn dialogue contexts grow long, models occasionally forget specific constraints and revert to verbose explanatory habits.

⚠️ Gotcha Warning [Rule Drift]: After exceeding 15 conversation turns, large language models tend to suffer from context fatigue. The workaround is resetting the session immediately after resolving a single task, or editing SKILL.md locally to reinforce critical constraint verbs.

⚠️ Gotcha Warning [Plugin Naming Collision]: Due to identical upstream and downstream repository names, installing directly via marketplace easily triggers local cache collisions. The workaround is strictly following official uninstall and cleanup commands to ensure only your active Fork source exists in the local registry.