keel-v2
Skill0 starsChanged 27 days agoMITView on GitHub
Keel's full rules, on demand: the eleven danger-surface definitions, the five invariants with their two enforcement layers each, the /review tripwires, and the full FAST/HEAVY router (one-shot authorization, context economy, priority over other skills, lanes, graphify-first comprehension). Load it before calling a task HEAVY or when any surface might be yes, before a HEAVY plan, before any /review pass, and whenever the CLAUDE.md Keel block's one-line summary is not enough to decide.
In this repository16 other files it gives its agents.
- skills/keel-v2/CLAUDE.mdCLAUDE.md
- doc-syncSkill
- fused-landingSkill
- keel-contentSkill
- keel-designSkill
- keel-devSkill
- keel-financeSkill
- keel-legalSkill
- keel-marketingSkill
- keel-opsSkill
- keel-qaSkill
- keel-researchSkill
- keel-secureSkill
- keel-setupSkill
- keel-verifySkill
- lazy-loopSkill
The file
Files
Keel v2 — the full rules (Tier 2: loaded on demand)
The project's CLAUDE.md carries only Keel's Tier-1 router: the lane rule, the
surface names, the rigor lines, the invariant names and an authorities index.
Everything that defines them lives in the two files next to this skill. Read
the one the task needs, in full; they are the rules verbatim, not a summary:
LAWS.md(this skill's directory) — the eleven danger surfaces with their definitions, the invariants with the two layers each one names, and the /review tripwires. Read it before a HEAVY call, a HEAVY plan, and every review.CLAUDE.md(this skill's directory) — the full router: §0 lanes and the pre-declaration rule, §1 FAST lane, §2 HEAVY lane, §3 one-shot authorization, §4 context economy and the memory protocol, §5 priority over other skills, §6 lanes, graphify-first comprehension and trade-off honesty. Where it says LAWS is "appended below this router", readLAWS.mdbeside it instead.
Precedence is unchanged: these rules outrank other skills; the project's Standing authorization block decides how FAST and HEAVY work end; a Keel gate (hook, deny rule, CODEOWNERS, ruleset) that blocks an action wins, and you take the path its message names.
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.

