T3 Tots skill
T3 Tots lets an agent read T3 Code thread state, answer questions and send reports, including to a phone
- 137 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Scripts connect to T3 Code's internal local API and read its database, so they see the content of agent threads
- Through `t3listen` agent questions and results go to a configured webhook, and approvals can be enabled for phone answers
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/t3-totsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Scripts connect to T3 Code's internal local API and read its database, so they see the content of agent threads
- Through `t3listen` agent questions and results go to a configured webhook, and approvals can be enabled for phone answers
This is third-party code. Review the repository files before installing.
What it does
A set of 22 skills and helper scripts for T3 Code, an editor with agent threads. The skills show thread status, list threads waiting for an answer, start new threads, fetch final output, build project reports and watch threads. They include `t3-status`, `t3-waiting`, `t3-watch`, `t3-new-thread`, `t3-digest` and `t3-cleanup`. Scripts ship alongside. `t3q.py` queries the T3 Code database read-only, `t3ctl` talks to the local API and installs skills, and the optional `t3listen` posts questions and run endings to your webhook. `manifest.json` records the size and SHA-256 of every file, and install checks them against the release tag.
Who it is for. People who run several agent threads in T3 Code and want to manage them from Grok Bot, Claude Code, Codex, Cursor or Hermes.
Good fit when
- Several threads run in parallel and you need a list of those waiting for your answer
- An agent finished a long run and you want the result on your phone through a webhook
- A thread stopped at a usage limit and should resume after the reset
- You want a weekly project report and a cleanup of old threads with a disk report
Not a fit when
- T3 Code is not installed: the skills work only with its local API and database
- You need a stable public interface: the T3 Code internal websocket may change after an update
- The official T3 Code MCP server for outside agents is enough
Example request
Show which T3 Code threads wait for my answer and briefly describe what each one asksLimitations
- The project is unofficial. `t3ctl` and `t3listen` use T3 Code's internal local websocket API, which is not public or stable, so a T3 Code update can break them.
- Permission approvals, live events to a webhook, resuming a stuck queue and usage-limit recovery are not yet covered by the official MCP server and the `t3 thread` command.
- Install from a release tag, not from `main`: every file is checked against `manifest.json`.
- On macOS a `t3ui` fallback drives the interface and needs Accessibility permission.
- Who may answer agent questions is set in the `t3-tots-rules` skill.
How to disable. Delete the skill folders from `~/.claude/skills`, `~/.agents/skills`, `~/.cursor/skills` or `~/.hermes/skills` along with the installed tools. Stop webhook delivery by disabling the `t3listen` login item.
FAQ
How do I install T3 Tots in Claude Code?
Download a release, unpack it and run `node tools/t3ctl skills install --to ~/.claude/skills`. Run with `--dry-run` first to see the file list.
What does t3listen send to the webhook?
The events `request.opened`, `request.resolved`, `thread.finished`, `thread.failed` and `thread.cancelled`. Approvals are sent only if you turn them on.
How do I filter out unneeded events?
Add a `filters` section to `listen.json` so run endings that need nothing from you are skipped. `thread.failed` and `request.opened` are never filtered.
Is there telemetry?
No, according to the author. The only outbound send is `t3listen` to a webhook you set up.
How is T3 Tots different from the official T3 Code MCP server?
The official MCP server and `t3 thread` do not yet cover approvals, live events and queue resume. The author moves features onto official interfaces as they gain coverage.
Discussion
No questions or reviews yet.
Related
SkillsA Go command-line client for Things 3 with a skill: reads the local task database, writes via URL scheme, undoes bulk edits.
The /myoffice command opens a browser office showing what every agent on your machine is doing: thinking, waiting for you or done.
Four skills for a person with several AI assistants: a task register, evidence-based acceptance, beginner instructions
tsk is a terminal task board: a TUI for you, a CLI and two skills for agents, and task launch in its own git worktree.
Official Todoist command line with an agent skill: tasks, projects and inbox from the terminal
Tokdash shows token usage and session cost for Claude Code, Codex, Gemini CLI and a dozen more agents locally
README in short
The README lists the repository contents in a table and describes the tools: `t3q.py`, `t3ctl`, `t3listen`, `thread-match.js`, export and cleanup scripts. Install from a release tag, phone answers and event filters follow.