kb
Skill0 starsChanged 4 days agoMITView on GitHub
Use when a KB hook message or check issue needs resolving, a source file is too large for one read, evidence or a table export arrives, the user asks whether the knowledge base already has a page or term for something, a stable technique may become a recipe, or the workspace needs an upgrade. Runs the integrity tool.
In this repository6 other files it gives its agents.
- add-moduleSkill
- checkSkill
- close-sessionSkill
- cross-module-owner-apiSkill
- curateSkill
- initSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Use this skill to act on what it reports and to do knowledge-base work through it instead of by hand.
- Very specific: line 17
Run from the workspace root (the folder holding AGENTS.md and modules.md):
- Very specific: line 24
Use python3 or py -3 where python is missing or older than 3.9.
- Distinctive: line 43
Read only the reference you need.
- Very specific: line 52
Never hand-edit what kb fix --safe would do.
- Very specific: line 53
When the user confirms that two pages are an intended duplicate, record it with kb config accept-duplicate <path> <path> so the issue does …
- Distinctive: line 59
After you edit a knowledge-base page or evidence file, the post-write hook applies safe fixes and reports remaining issues for that file.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
KB
The workspace ships a deterministic integrity tool in tools/kb/. It
checks links, indexes, evidence tracking, required page sections,
duplicates, recipes, and encoding; it applies fixes that have exactly
one correct answer and hands everything else to you as numbered issues.
Use this skill to act on what it reports and to do knowledge-base work
through it instead of by hand.
How to call it
Run from the workspace root (the folder holding AGENTS.md and
modules.md):
python tools/kb/kb.py <command> ...Use python3 or py -3 where python is
missing or older than 3.9. Every command prints a one-line
summary first. Exit 0 means clean or done; 1 means issues were found;
2 means a usage or workspace error; 3 means a table export needs a new
layout; 4 means an upgrade has open questions.
Route by situation
| Situation | Read |
| A source file is larger than about 50 KB, or you only need its shape | references/outline.md |
| New evidence arrived (documents, exports, library sources, examples) | references/ingest.md |
| A table export must become table pages | references/tables.md |
| You are about to write a page and need to know whether one exists | references/dedupe.md |
| An unknown term, abbreviation, or transliteration | references/terms.md |
| A platform technique has stabilized and may be reused | references/recipes.md |
A hook message, a check result, or docs/checks/OPEN.md lists issues | references/issues.md |
| The session-start line says the workspace version is old | references/upgrade.md |
Read only the reference you need.
Fix policy
- Safe fixes (exactly one correct answer) are applied automatically by
the hooks and by
kb fix --safe; each one is journaled and can be reverted withkb undo <run-id>. - Ambiguous cases become issues with their candidates. You decide when the evidence settles it; otherwise ask the user.
- Never hand-edit what
kb fix --safewould do. Run the command. - When the user confirms that two pages are an intended duplicate,
record it with
kb config accept-duplicate <path> <path>so the issue does not come back.
What the hooks do
After you edit a knowledge-base page or evidence file, the post-write
hook applies safe fixes and reports remaining issues for that file. At
the end of a turn, the stop hook blocks once if your changes introduced
a blocking issue. At session start, one line reports open issues or an
old workspace version. When nothing is wrong, the hooks print nothing.
Without hooks, run kb check --changed <files> yourself before closing
a session.
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.