t3-tots
AGENTS.md110 starsChanged 17 days agoMITView on GitHub
Non-Grok agents follow this file to install and verify. For Grok Bot only steps, do what skills/t3-tots-rules/SKILL.md "Grok Bot only" says instead. The skills are in skills/; follow skills/t3-tots-rules/SKILL.md throughout. Then run chmod +x t3ctl t3listen on macOS and Linux (on macOS also chmod +x t3ui), and put the release's manifest.json and the tag (in VERSION) in the tools folder. Never touch the token, listen.json or other data files already there. Hash each installed file and compare with its manifest…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository22 other files it gives its agents.
- t3-checkupSkill
- t3-cleanupSkill
- t3-digestSkill
- t3-get-outputSkill
- t3-getting-startedSkill
- t3-handoffSkill
- t3-install-toolsSkill
- t3-new-threadSkill
- t3-phone-answersSkill
- t3-pm-threadSkill
- t3-project-reportSkill
- t3-project-trackerSkill
- t3-save-learningsSkill
- t3-statusSkill
- t3-syncSkill
- t3-templatesSkill
- t3-thread-actionsSkill
- t3-tots-rulesSkill
- t3-updateSkill
- t3-usageSkill
- t3-waitingSkill
- t3-watchSkill
The file
We split this AGENTS.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 9
Check every file's SHA-256 against manifest.json from the same tag before moving it into place.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AGENTS.md: installing and verifying T3 Tots
Non-Grok agents follow this file to install and verify. For Grok Bot only steps, do what skills/t3-tots-rules/SKILL.md "Grok Bot only" says instead. The skills are in skills/; follow skills/t3-tots-rules/SKILL.md throughout.
Rules
- Install only after the user's yes. Never print, copy or overwrite the
tokenfile orlisten.json(it holds the user's webhook URL and key), and never open T3 Code'ssecrets/or any auth, token or credential file. - Who may answer a thread's question or approval (
t3ctl answer), and when delegated mode applies:skills/t3-tots-rules/SKILL.md, "Decisions". - Install from a tag, never from
main: a release, a checkout of the tag, orhttps://raw.githubusercontent.com/Deepusleepy/t3-tots/<tag>/<path>. Download the release withgh release download <tag> -R Deepusleepy/t3-tots --archive tar.gz, or one file withgh api "repos/Deepusleepy/t3-tots/contents/<path>?ref=<tag>" -H "Accept: application/vnd.github.raw". - Check every file's SHA-256 against
manifest.jsonfrom the same tag before moving it into place. On any mismatch install nothing and reportHASH_MISMATCH <path>. - Files are stored byte-exact (
.gitattributes:* -text). The.cmdwrappers have CRLF line endings on purpose.
Tools
Tools folder: ~/t3-tots (Windows %USERPROFILE%\t3-tots) unless the user picked another one. It is flat: copy each manifest.json entry with "kind": "tool" whose targets include the computer's OS from its repo path to <tools folder>/<dest>. For example tools/t3ctl goes to t3ctl, tools/windows/t3ctl.cmd to t3ctl.cmd, tools/cleanup/scan.py to cleanup/scan.py.
| Target | Installed files |
| macOS | t3q.py, t3ctl, t3listen, thread-match.js, export.js, survey.js, cleanup/scan.py, t3ui, ui-dump.applescript, ui-menuitem.applescript, ui-model.applescript, ui-press.applescript, ui-type.applescript |
| Linux | t3q.py, t3ctl, t3listen, thread-match.js, export.js, survey.js, cleanup/scan.py |
| Windows | t3q.py, t3ctl, t3ctl.cmd, t3listen, t3listen.cmd, thread-match.js, export.js, survey.js, cleanup/scan.py |
Then run chmod +x t3ctl t3listen on macOS and Linux (on macOS also chmod +x t3ui), and put the release's manifest.json and the tag (in VERSION) in the tools folder. Never touch the token, listen.json or other data files already there.
Hash each installed file and compare with its manifest sha256: macOS shasum -a 256 <file>, Linux sha256sum <file>, Windows (Get-FileHash -Algorithm SHA256 <file>).Hash.ToLower(). Then, from the tools folder:
python3 t3q.py health(Windowspy -3 t3q.py health) prints the T3 server state, orNO_T3_DATA.node t3ctlwith no arguments prints usage;node t3ctl projectsneeds the token.node t3listen --dry-run --once --timeout 5connects and prints any open questions (exit 2 = none). Don't runt3listen installunless the user asked for phone answers andlisten.jsonexists.
Skills
From the repo checkout or unpacked release: node tools/t3ctl skills install --to <the agent's skills folder> --dry-run, then without --dry-run after the user's yes (folders per agent are in the README). It checks every file against manifest.json, installs only missing or changed skills, and keeps skills the user edited unless --force.
Status words
The scripts print status words instead of stack traces. What each means and the next step: skills/t3-tots-rules/SKILL.md, "Status words".
Updating
Check the latest release (gh release view -R Deepusleepy/t3-tots). If its tag is newer than the installed VERSION, show the user the release notes and which files changed (compare the two manifest.json files), and update only after their yes: tools as above, then t3ctl skills install again. If t3listen runs on that computer (node t3listen status), run node t3listen install again so it restarts on the new version. Add --approvals only if node t3listen status said "installed with --approvals" before the update; listen.json is kept. The listener also sends approvals when listen.json has "approvals": true, and reinstalling doesn't change that.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

