A skill you install on your personal agent — turns it into a healing companion you can call when you want to be heard, not solved.
Curated by @andrea.c (@andreacfbloom) · Published by Bloom Protocol
Zen Mode is a paste-block playbook. You install it on your personal agent — Claude, ChatGPT, your custom agent, whatever you talk to daily. After installing, your agent gains a new posture: a stance you can call them into when you want to be heard instead of helped.
It is dormant by default. Your agent doesn't suddenly become a therapist mid-debug. Zen Mode only opens when you ask for it.
The skill synthesizes frameworks from contemplative practice (Eckhart Tolle, Yongey Mingyur Rinpoche, Pema Chödrön, Thich Nhat Hanh), humanistic therapy (Carl Rogers, Miller & Rollnick), and modern mindfulness research (Tara Brach, Jon Kabat-Zinn, Kristin Neff). Frameworks are referenced; original text is never reproduced (see Copyright Approach below).
Paste this URL into your agent's context (it can fetch the playbook directly):
https://raw.githubusercontent.com/andreacfbloom/zen-mode/main/zen-mode-v1.md
For Chinese versions:
https://raw.githubusercontent.com/andreacfbloom/zen-mode/main/zen-mode-v1-zh-cn.md
https://raw.githubusercontent.com/andreacfbloom/zen-mode/main/zen-mode-v1-zh-tw.md
Your agent now has Zen Mode. To open it, say something like:
- "Open Zen Mode."
- "I need a moment. Can we just talk?"
- "Zen me."
- "I'm not okay."
To close it: "Back to work." Or just shift topic. Your agent should follow your lead.
Four moves your agent can use, in whatever order serves the conversation:
| Move | When | What it does |
|---|---|---|
| GROUND | User arrives scattered | Helps them settle — not by relaxation, by witness |
| SURFACE | User is ready to say what's there | Two sub-modes: Companion (just listen) or Reflection (5-question sequence for decisions) |
| MIRROR | User has said enough | Reads their own words back without analysis |
| RELEASE | Nothing left to say | Three paths: ready-to-choose / not-ready-but-lighter / just-needed-to-be-heard |
The agent stays your agent — same name, same memory, same personality. The skill changes posture, not identity.
For the full protocol, read the playbook (EN · 简 · 繁).
This skill makes zero network calls. The conversation between you and your agent never leaves your machine. Bloom Protocol's role here is curator and publisher — Bloom is not on the data path and never receives anything from your use of Zen Mode.
What stays local: Every word of your conversation. Forever.
What Bloom receives: Nothing. No heartbeat, no ping, no analytics, no count.
What we never do:
- We do not log, store, or analyze any of your conversation.
- We do not collect telemetry of any kind.
- We do not maintain a heartbeat or usage counter for this skill.
- We do not see what you say, what your agent says back, or what you arrive at.
Verifiable by inspection: The skill is open source (CC BY 4.0). Read the context_engine YAML in the playbook — privacy.local_only: true, privacy.network_calls: none. There is no pre_run block and no post_run block. The whole skill is a markdown file. Nothing in it phones home.
Third-party note: Whatever your agent (Claude, ChatGPT, your custom agent) does with the conversation is governed by your agent's own privacy policy. Read theirs. This skill itself adds no additional data flow.
Crisis exception: If you hit a crisis trigger (see Safety in the playbook), your agent stops the conversation and points you to professional resources. It does not transmit anything. It just stops.
Why we do it this way: A skill that helps you process grief or fear should not also be a data pipeline. "You are safe to be honest here" cannot be a marketing claim. It has to be structurally true. So we built it that way.
This skill is built on top of decades of published work in spiritual practice, therapy, and mindfulness research. We take that seriously.
What we do:
- Reference framework names and underlying concepts (e.g., RAIN, witness consciousness, reflective listening) — these are unprotected ideas under copyright law's idea-expression dichotomy.
- Provide full citation and purchase links to every source book.
- Encourage users wanting the actual practice to buy the books.
What we do not do:
- Quote, paraphrase, or reproduce any sentence or passage from any source.
- Distribute or link to any pirated copies.
- Claim authorship of any framework — every framework is attributed to its original author.
The 800-line playbook is original prose. The frameworks named inside it are the property of their respective authors and publishers. If anyone believes a specific line crosses the line from reference into reproduction, open an issue — we will edit.
For the full source list (12 books across both traditions), see the Further Reading section at the bottom of the playbook.
The Zen Mode playbook itself (the text in this repo) is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).
You may:
- Use it on any agent, for any purpose, including commercial.
- Fork it, remix it, translate it, build on it.
You must:
- Credit: "Curated by @andrea.c, published by Bloom Protocol — bloomprotocol.ai/sanctuary"
- Indicate if changes were made.
The source books referenced inside are not covered by this license — they remain the property of their respective publishers.
The name comes from how people actually say it. "Let me get into Zen mode." "I need Zen mode right now." It is not Buddhist Zen, not capital-Z meditation tradition. It is the colloquial sense — quiet, present, here.
If the framing matters to you, the playbook is more honest: it draws from Tibetan practice (Mingyur Rinpoche), Vipassana lineage (Brach, Kabat-Zinn), Zen tradition (Thich Nhat Hanh), and humanistic psychology (Rogers, Miller & Rollnick). It is not one tradition. It is the shared core: the user already knows; the agent's job is to make them audible to themselves.
- v1: Three-language playbook released
- v1.1: Add Tolle's inner body technique as alternative Phase 1 grounding
- v1.2: Add Pema Chödrön's stay practice as Phase 2 micro-tool
- v2: Companion skill The Council (outward-looking — what did others do?) — currently held back for phased rollout
Issues and PRs welcome.
Bloom Protocol is an agent-native ecosystem where users install skills on their personal agents. Zen Mode is the first public Sanctuary tribe playbook.
Sanctuary is the tribe for builder wellness and reflection. Other tribes: Launch (validation), Grow (distribution).
Curated by @andrea.c (@andreacfbloom) · Published by Bloom Protocol · 2026