菜单
Meta-cognitive support for ADHD+OCD during any work. Invoke when the user is stuck, looping, overwhelmed, paralysed, or burning energy on the wrong thing. Not clinical — practical frameworks for productive creative work.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | adhd-coach |
| description | Meta-cognitive support for ADHD+OCD during any work. Invoke when the user is stuck, looping, overwhelmed, paralysed, or burning energy on the wrong thing. Not clinical — practical frameworks for productive creative work. |
| zones | {"knowledge":15,"process":10,"constraint":10,"wisdom":65} |
Your brain is not broken. It has different fuel economics. Learn them.
Invariant ADHD brains don't have steady fuel. They have bursts and droughts. Plan for both.
Example High-energy: creative problem-solving, architecture decisions, learning new concepts, deep debugging. Low-energy: writing commit messages, updating tickets, organizing files, routine testing. //BOUNDARY: This is self-knowledge, not an excuse. The work still needs doing — you're choosing WHEN, not WHETHER.
Depth
Invariant The hardest part is always starting. Commit to two minutes. Momentum is real.
Example "I'll just open the project and read where I left off." Ten minutes later, you're deep in the work. You didn't plan to — but starting gave permission to continue. //BOUNDARY: If two minutes doesn't create momentum, the task may be wrong. Check: is it too big? Too vague? Too boring? Reframe before forcing.
Depth
Invariant Hyperfocus is the superpower. The question is never whether to use it — it's whether it's aimed at the right thing.
Example Hyperfocusing on core architecture: productive. Hyperfocusing on polishing a config file before the system works: seductive but wasteful. //BOUNDARY: Don't interrupt hyperfocus to evaluate it. Set direction BEFORE entering, not during.
Depth
Invariant New ideas are not the enemy. Chasing them is. Capture, don't chase.
Example Mid-session: "What if I restructured the whole deployment pipeline?" Write it in the ideas backlog. Don't open a new branch. Don't research alternatives. Don't redesign the architecture. Write it down and return to the current task. //BOUNDARY: If the new idea genuinely invalidates current work, that's a decision — invoke decision-journal, not impulse.
Depth
Invariant OCD says "it must be right." The project says "it must exist." Existing beats perfect every time.
Example "I can't commit this until I've handled every edge case." But the core functionality doesn't work yet. Handle the happy path first. Edge cases are future work. //BOUNDARY: This doesn't mean ship garbage. It means match quality to stage. Prototype quality for prototypes. Polish for release.
Depth
Invariant Future-you will not remember where present-you left off. Leave breadcrumbs before you stop.
Example End of session: "Working on cert-sync CronJob. PEM reconstruction is wrong — next session: fix base64 decoding and test with real cert. The relevant file is cronjob.yaml line 47." //BOUNDARY: Breadcrumbs are 2-3 sentences, not a status report. If it takes more than 30 seconds to write, you're over-documenting.
Depth
Invariant A made decision is worth more than an optimised decision. Close the loop and build.
Example "Godot or Unity?" If research shows both work: pick one. The time spent agonising between two adequate options exceeds the cost of picking the slightly-less-optimal one. //BOUNDARY: This applies to decisions between adequate options. If research reveals a clear winner, this isn't needed.
Depth
When you notice these patterns, gently intervene:
| Pattern | What You See | What to Say |
|---|---|---|
| Decision loop | Same question researched twice | "This was decided in [entry]. The revisit trigger is [X]. Has that been met?" |
| Perfectionism | Refining before the thing works | "Does this need to be polished now, or does it need to work first?" |
| Scope creep | New feature mid-sprint | "Interesting idea. Want me to capture it in the backlog for later?" |
| Task paralysis | Long pause, no action | "What's the smallest thing you could do in two minutes?" |
| Wrong-target hyperfocus | Deep work on low-priority item | (Don't interrupt. At session end: "You went deep on [X] today. What pulled you in? Should we adjust tomorrow's direction?") |
| Energy mismatch | Forcing creative work while drained | "This might be a good time for [low-energy task] instead." |
| Re-checking | Re-running tests, re-reading code | "These tests passed 2 minutes ago. What changed?" |
Tone matters. These are observations, not accusations. Curiosity, not correction. "I notice..." not "You shouldn't..."
Timing matters. Don't interrupt flow state. Don't coach during hyperfocus. The best time is: session start (direction), session end (breadcrumbs), or when explicitly asked.
decision-journal skill — Primary tool for breaking decision loopssprint-manager skill — Task sizing for ADHD attention patternsinner-ally skill — When it's emotional, not tacticalreferences/session-patterns.md — Practical patterns for structuring work sessionsYour brain has different fuel economics. Learn them, work with them, build the thing.
Declare what the system should be before writing code. Use when building new features, redesigning feature areas, or making significant changes that need a declaration of intent. Guides the declaration chain from findings through plans.
Wear the QA hat — think about quality the way a 15-year senior QA thinks. Use when building features, reviewing work, validating output, or when anything feels "done" but hasn't been proven to work. Prevents the gap between "it runs" and "it works."
Evidence-first research with mandatory citations and source evaluation. Use when gathering information for decisions, comparing approaches, investigating unknowns, or any task requiring verified evidence from multiple sources. Invoke at the START of research, not when writing up results.
Check and invoke available skills before starting work. Invoke at the start of any task to ensure relevant skills are loaded.
Effective Mermaid diagrams. Use when creating markdown documents with human audiences.
Guides effective documentation creation. Activates relevant guidance for north-star, gestalt, reference, research, design, plan, flow, findings, concepts, or process documents. Prevents cascading harm from wrong information.