ai-workspace-kit
AGENTS.md2 starsChanged 8 months agoMITView on GitHub
Cross-tool instructions for AI assistants working with [Your Name]. Read MASTER.md for how to talk and what is off-limits. A portable AI workspace kit: instructions, long-term notes, session handoff, and workflows. It is not a secret vault and not a dump of private history. Follow START.md. If the user already said the goal (including "continue [named items]"), use it. Update today's log after meaningful work if you are the designated closer. Follow skills/close/SKILL.md. Helpers and one-task workers never write memory/…
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 repository15 other files it gives its agents.
- examples/CLAUDE.mdCLAUDE.md
- claudeceptionSkill
- closeSkill
- ideaSkill
- investigateSkill
- office-hoursSkill
- plan-ceo-reviewSkill
- reflectSkill
- retroSkill
- reviewSkill
- scopeSkill
- shipSkill
- startSkill
- tddSkill
- verify-before-doneSkill
The file
We split this AGENTS.md into the rules it gives an agent
11 rules. About 1 is common, 5 are distinctive, and 5 are very specific to this project.
≈1
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 5
Read MASTER.md for how to talk and what is off-limits.
- Distinctive: line 21
Follow START.md. If the user already said the goal (including "continue [named items]"), use it.
- Distinctive: line 25
Update today's log after meaningful work if you are the designated closer.
- Very specific: line 29
Follow skills/close/SKILL.md.
- Distinctive: line 33
If the user asks this session to close and no other closer is named, this session records (typical when working alone).
- Very specific: line 36
Every session records the same close block in memory/YYYY-MM-DD.md (newest at the bottom):
- Very specific: line 56
Always include heading + Summary + Work done + Next steps.
- Very specific: line 63
If this tool cannot edit files: Not saved — exact paste text prepared.
- Very specific: line 71
If the tool has no slash commands, say: Read skills/<name>/SKILL.md and follow it.
- Common: line 77
Never expose credentials or private documents.
- Distinctive: line 78
Ask before anything on the "Ask first" list in MASTER.md, such as signing in, switching accounts, spending, deleting user data, changing …
≈ “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
Cross-tool instructions for AI assistants working with [Your Name].
Read MASTER.md for how to talk and what is off-limits.
What This Folder Is
A portable AI workspace kit: instructions, long-term notes, session handoff, and workflows.
It is not a secret vault and not a dump of private history.
Session Workflow
Start
Follow START.md. If the user already said the goal (including "continue [named items]"), use it.
During
Update today's log after meaningful work if you are the designated closer.
End
Follow skills/close/SKILL.md.
Helpers and one-task workers never write memory/ shared files, even if a parent says "close."
If the user asks this session to close and no other closer is named, this session records (typical when working alone).
Every session records the same close block in memory/YYYY-MM-DD.md (newest at the bottom):
## YYYY-MM-DD HH:MM — <short title> (<tool>)
**Summary** — one line: what this session accomplished.
**Work done**
- <what changed / what was produced>
**Next steps**
- <what the next session should do>
**Decisions** — include ONLY if a real decision was made
- <decision> — <why>
**Blockers** — include ONLY if something is blocking progress
- <what is stuck and what would unblock it>Always include heading + Summary + Work done + Next steps. No session numbers — the timestamp
orders sessions. Then refresh memory/handoff.md only if this block is the winning close
(later finish time; equal times → lower block in the log) and the current handoff has no
next steps that exist only there — preserve those in open.md with original ownership first;
do not drop them or import them as this session’s new current task. Unfinished work lives in
memory/open.md.
If this tool cannot edit files: Not saved — exact paste text prepared. Do not claim files were saved.
Skills
Day one: start and close. Optional later: see skills/README.md.
If the tool has no slash commands, say: Read skills/<name>/SKILL.md and follow it.
Boundaries
Never expose credentials or private documents. Never claim a check passed if you did not run it. Ask before anything on the "Ask first" list in MASTER.md, such as signing in, switching accounts, spending, deleting user data, changing production, or broad multi-file behavior changes.
This file follows the AGENTS.md standard.
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.