lore-skill
AGENTS.md0 starsChanged 34 days agoMITView on GitHub
This project uses KIS memory under kis/. Knowledge = stable facts. Intent = goals and plans. State = current reality. - Read kis/state/ before planning or implementing, then only the Intent and Knowledge the task needs. - Put each fact in exactly one layer, and update an existing file instead of creating a new one. - Prove work with real command or verification output before marking anything done. - Synchronize the KIS layers that changed when work finishes. - Full…
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 32,671 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.
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 6
Read kis/state/ before planning or implementing, then only the Intent and Knowledge the task needs.
- Distinctive: line 7
Put each fact in exactly one layer, and update an existing file instead of creating a new one.
- Very specific: line 45
Keep SKILL.md short; it is loaded into context on every use.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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
<!-- kis:anchor:start -->KIS Project Memory
This project uses KIS memory under kis/. Knowledge = stable facts. Intent = goals and plans. State = current reality.
- Read
kis/state/before planning or implementing, then only the Intent and Knowledge the task needs. - Put each fact in exactly one layer, and update an existing file instead of creating a new one.
- Prove work with real command or verification output before marking anything done.
- Synchronize the KIS layers that changed when work finishes.
- Full instructions:
.agents/skills/kis/SKILL.md. - Commands live in
.agents/skills/kis/commands/: start, plan, act, sync, check, init. Claude/kis:start, Pi and OpenCode/kis-start, Codex/prompts:kis-start. Antigravity has no custom slash commands: name the step and follow its command file.
<!-- kis:anchor:end -->Agent Instructions
This repository is the source of lore. Nothing here is a live install: skills/ sits
outside every host's discovery path. To use it, install into a home with bootstrap.sh.
Layout
skills/lore/ the lore package, and what ships
SKILL.md the method; loaded at work time
commands/ sync, learn-taste, retro (Claude: /lore:sync, /lore:learn-taste, /lore:retro)
hosts/ one adapter per host: global file, hooks, transcripts, model commands
scripts/ lore.py (store, index, delivery, import), lore_capture.py (hooks,
harvest, review, taste), lore_sync.py (hub sync), lore-hook.sh
scripts/ install-skill.sh, run from a source clone only
docs/ the pages the README links to, and its SVG images (never installed)
.github/workflows/ CI: tests/run.sh on Linux
tests/ one suite per area (store, delivery, capture, review, sync, import, taste,
setup, review fixes, install), never installed
bootstrap.sh one-command install, check, update
kis/ this repo's own project memory (read kis/state first)Installing copies skills/lore/ verbatim to ~/.agents/skills/lore and links it for Claude
and agy. If a file does not belong in someone's home, it does not belong in
skills/.
Rules
- Keep
SKILL.mdshort; it is loaded into context on every use. Detail goes in references. SKILL.mdis instructions at work time; it never documents this repo or the installer.- Command bodies in
skills/lore/commands/stand alone (loaded withoutSKILL.md), so they restate rules on purpose. When a command andSKILL.mddisagree,SKILL.mdis right. metadata.versioninSKILL.mdis bumped on every change that ships.- Tests never call a real model CLI: they set
LORE_EXTRACT_CMDandLORE_REVIEW_CMDto fakes. - Guards live in
skills/lore/scripts/andscripts/install-skill.sh; every one is mutation-checked (break it, watch its test fail, restore it). - The live install is the owner's working copy: after a change here, run
bootstrap.sh update --source .and prove it withcheck.
Checks
tests/run.shDoes 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.

