Skip to content
Back to search

phoenix-protocol

MarketplaceWorks with Claude CodeGitHub

Rebuild a daily rhythm after a breakup, lost job or burnout: one small action from today, evening check-ins, a minimal day after a miss, weekly reviews. Numbers by a local script; files stay on your computer.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/tsirbiladze--phoenix-protocol/badge.svg)](/listings/tsirbiladze--phoenix-protocol)

HTML

<a href="/listings/tsirbiladze--phoenix-protocol"><img src="/api/listings/tsirbiladze--phoenix-protocol/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

<details open> <summary><b>Any agent</b> (Claude Code, Codex, Cursor, Gemini CLI, …)</summary>

npx skills@latest add tsirbiladze/phoenix-protocol

</details>

<details> <summary><b>Claude Code plugin</b></summary>

/plugin marketplace add tsirbiladze/phoenix-protocol
/plugin install phoenix-protocol@phoenix-protocol

</details>

Then ask your agent: *"Help me get back on track."* Later, *"check-in"* or *"weekly review"*.

It needs an agent that works with files on your computer and can run Python 3.9 or newer. Nothing else is installed.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 15 lines
  1. {
  2. "name": "phoenix-protocol",
  3. "owner": { "name": "Tornike Sirbiladze", "url": "https://tsirbiladze.com" },
  4. "metadata": {
  5. "description": "Rebuild a daily rhythm after a hard stretch, with your AI agent."
  6. },
  7. "plugins": [
  8. {
  9. "name": "phoenix-protocol",
  10. "source": "./",
  11. "description": "Rebuild a daily rhythm after a breakup, lost job or burnout: one small action from today, evening check-ins, a minimal day after a miss, weekly reviews. Numbers by a local script; files stay on your computer.",
  12. "category": "productivity"
  13. }
  14. ]
  15. }
README.md · README · 66 lines
  1. # Phoenix Protocol
  2. An agent skill that helps you get your days back after a hard stretch: a breakup, a lost job, burnout, a move, a bad season. You start today with one small action and an evening check-in. Your agent comes by every evening. The plan grows only when a real week shows it can. A missed day has a written answer, and that answer is never catching up.
  3. Most plans are made for the perfect day. This one is made for the day after you fail.
  4. ![The dashboard of a fictional person: the week's score, days kept, returns, the cycle's weeks and days, today's plan](docs/dashboard.png)
  5. The full description, with what it does not do: https://tsirbiladze.com/skills/phoenix-protocol
  6. ## Install
  7. <details open>
  8. <summary><b>Any agent</b> (Claude Code, Codex, Cursor, Gemini CLI, …)</summary>
  9. ```bash
  10. npx skills@latest add tsirbiladze/phoenix-protocol
  11. ```
  12. </details>
  13. <details>
  14. <summary><b>Claude Code plugin</b></summary>
  15. ```
  16. /plugin marketplace add tsirbiladze/phoenix-protocol
  17. /plugin install phoenix-protocol@phoenix-protocol
  18. ```
  19. </details>
  20. Then ask your agent: *"Help me get back on track."* Later, *"check-in"* or *"weekly review"*.
  21. It needs an agent that works with files on your computer and can run Python 3.9 or newer. Nothing else is installed.
  22. ## How it goes
  23. 1. **Start** (ten minutes, one question at a time): how you are, your country (for the right emergency numbers), what happened in one word or none, one anchor action written as "when → then", a minimal day for bad days, and an evening check-in time. Your first check-in is the same evening.
  24. 2. **Every evening**, two to five minutes: what got done, in one tap; one optional question; tomorrow's first step.
  25. 3. **After a miss**: one accepting line, then one small action. Two heavy days make the next one a minimal day. After three silent days you get a welcome back with no numbers. There is no "from Monday".
  26. 4. **Every seventh day**, a weekly review: what happened, one obstacle and its fix, three actions for next week. The script decides the level: 80% or more adds one item, under 50% makes the next week lighter, and nothing ever goes back to zero.
  27. Opt-in modules: an impulse pause for the urge to text or check on someone, a worry window, and an idea bank.
  28. ## Safety
  29. At the start of every session, the agent says it is an AI helper, not a therapist and not an emergency service. Crisis signs stop the routine. The agent asks one plain question and gives a crisis line for your country from a dated table checked against official sources (`skills/phoenix-protocol/scripts/helplines.json`). When the danger is immediate, the emergency number comes first. The agent refuses to help with tracking a person, pressuring someone back, or contact after a block or a request for no contact. The rules are in [`references/safety.md`](skills/phoenix-protocol/references/safety.md).
  30. ## Privacy
  31. Your plan is a folder on your computer (`~/phoenix-protocol` by default) of plain JSON, CSV and Markdown files. Git keeps a local history if you want one. The skill never adds a remote or pushes on its own. Unsent texts and parked worries stay in `private/`, outside git. Other people appear by role, not by name, and nothing about your health is written down. The scripts make no network calls. What you write to your agent goes to the model provider like any chat.
  32. ## Limits
  33. Local coding agents only: claude.ai and ChatGPT keep no folder between sessions. The plan file and the dashboard are in English or Georgian, and your agent speaks your language. One person per folder. Not for anyone under 18, and not therapy. The 80% and 50% thresholds and the 12-week cycle are conventions, not findings, and you can change them.
  34. ## In this repository
  35. | Path | What it is |
  36. |---|---|
  37. | `skills/phoenix-protocol/` | The skill as it is installed: `SKILL.md`, one reference per moment in `references/`, and `scripts/`: `phoenix.py` (every number, date and rule), `dashboard.py`, `strings.json` (every sentence for the person, in English and Georgian), `helplines.json` |
  38. | `docs/` | The skill's page on tsirbiladze.com in English and Georgian, the dashboard pictures, and `research.md`: the evidence behind the design |
  39. | `tests/` | Unit tests of the scripts and the docs (`python3 -m unittest discover -s tests -t .`), three fictional people run through scripted weeks, and `tests/rehearsal/`: how a fresh agent rehearses the whole skill |
  40. | `SPEC.md` | What the skill must do, rule by rule |
  41. ## License
  42. MIT

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.