softcane

softcane / human-state-skills

Public

Small response modes for coding when your head is in a different state.

adhd-friendly ai-agents burnout claude-skills codex-skill
17
0
100% credibility
Found May 24, 2026 at 29 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
AI Summary

human-state-skills is a collection of special response modes for AI coding assistants that help when you're working in difficult mental states. Instead of giving the same detailed response regardless of how you feel, the modes adapt to situations like being overwhelmed during an incident, too foggy to think clearly, burned out and avoiding work, or stuck in a long debugging loop. The project includes seven modes: overloaded-mode for crisis situations, foggy-mode for when you can't think straight, plan-compass for breaking big decisions into tiny pieces, reality-check-mode for stepping back from AI rabbit holes, burnout-mode for when exhaustion is blocking action, brain-fog-mode for minimal guidance, and normal-mode to return to standard responses. It's designed as a plugin for Claude Code and Codex AI assistants.

How It Works

1
💻 You start using an AI coding assistant

You discover you can talk to an AI to help with coding tasks and it works surprisingly well.

2
😰 The AI doesn't understand when you're overwhelmed

When you're stressed or foggy, the AI keeps giving you long lists of things to do, making everything worse.

3
🔍 You find human-state-skills

Someone shares a collection of special modes that help the AI respond to your mental state instead of just the task.

4
⚡ You add the modes to your AI assistant

You install the skill pack so your AI knows how to help you when you're overloaded, foggy, or burned out.

5
You pick the mode that fits how you feel
🚨
Overloaded mode

When everything is on fire, the AI gives you one priority and tells you what to drop.

🌫️
Foggy mode

When you can't think straight, the AI gives you one tiny step at a time.

🧭
Plan compass

When planning feels overwhelming, the AI asks one decision at a time.

🛑
Reality check mode

When you've been debugging too long, the AI gently brings you back to reality.

✨ You get help that actually fits

The AI responds to your mental state, not just the technical problem, so you can move forward no matter how you feel.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 29 to 17 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is human-state-skills?

This is a collection of prompt templates for AI coding assistants that adapt their responses based on your mental state. Instead of getting the same verbose technical answer whether you're sharp at 10am or foggy after a bad deploy, these modes shape how the AI responds. The package includes modes for overloaded (one priority, do/defer/drop), foggy (tiny steps, state kept external), plan-compass (stress-testing decisions one at a time), reality-check (grounding when you've been in an AI loop too long), and burnout (breaking the shame loop before planning). It installs as a plugin for Claude Code or as skills for Codex.

Why is it gaining traction?

The insight is sharp: your cognitive state changes how useful an AI response is. A normal response during on-call overload just adds noise. These modes force the AI to match your capacity. The before/after examples are the real sell--seeing a 9-step response collapse into "roll back, post one message, stop" makes the value immediate. Developers are hungry for ways to make AI tools actually useful in high-pressure moments, not just during clean green-field coding sessions.

Who should use this?

On-call engineers who need fast, constrained responses during incidents. Developers working through burnout or brain fog who need the AI to hold state for them. Anyone who has spent too long in an AI debugging loop and needs grounding. If you use Claude Code or Codex daily, this is worth a look.

Verdict

The concept is genuinely useful and the examples land well. At 17 stars and a 1.0% credibility score, this is early-stage and unproven at scale. No tests, thin documentation beyond the README. Worth installing if you want to experiment with state-aware AI interactions, but don't bet production workflows on it yet. Watch for community adoption before committing.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.