greenroomSAFE
Keep hundreds of Claude Code skills off the context window without ever losing them: hidden skills, one gate per family, a session map, zero maintenance.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Keep hundreds of Claude Code skills off the context window without ever losing them.
Español
In a theater, the green room is where performers wait off stage: not taking up the stage, but ready, and the stage manager knows who is there and calls them on when it's their cue. greenroom does that for Claude Code skills.
Quick start
git clone https://github.com/PapiScholz/greenroom cd greenroom python install.py --dry-run # 1. preview: which family each of your skills would go to (writes nothing) python install.py # 2. install (asks before hiding anything)
Then:
- Open a new Claude Code session. The model now starts with a short skill map: the gates that exist and the ones the current repo needs.
- Keep working as usual. Install skills with
npx skills add ...as always. greenroom files each one under its family, hides it and lists it in the right gate on its own. - Ask for anything. The model enters the matching gate (
gate-motion,gate-data...), which loads only the skills the task needs. You can also type/gate-motion, or any hidden skill's/name: they are all still in your menu.
Forgot how it works? Run this:
python ~/.claude/greenroom/greenroom.py status
The problem
Every skill you install puts its name and description in the context of every session. Past a few dozen skills that gets expensive. Past ~1% of the context window, Claude Code drops the descriptions of the skills you use least, so the model picks worse or never finds the skill that would have done the job right.
The usual fixes make it worse:
- Uninstalling loses the skill.
- `skillOverrides: "off"` saves the context, but it also removes
/namefrom your menu. Nothing tells the model the skill exists, so it improvises instead of using it.
How it works
flowchart TD I["Yo
b10433312d68OBSERVED · 2026-10-10Install
Commands as the repository documents them. They are shown, not run.
git clone https://github.com/PapiScholz/greenroom
git clone https://github.com/PapiScholz/greenroom
Host compatibility
What the documentation claims. We have not run a compatibility test.
| Host | Status | Notes |
|---|---|---|
| claude-code | mentioned |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (1 observation(s))
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (1)
Gates applied: no_behavioural_pass.
b10433312d68full audit observations/trust-audit/skill/papischolz__greenroom.json · Report an issue / request a re-scanAudit history
Every audit this skill has had.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-10 | b10433312d68 | SAFE | B | 89 | first audit |
Questions
What does the greenroom skill do?
Keep hundreds of Claude Code skills off the context window without ever losing them: hidden skills, one gate per family, a session map, zero maintenance.
Is greenroom safe to install?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean skill reads B.
What can greenroom access on my machine?
The audit observed that it reads or writes files. Each of those is consistent with what it says it does. Secrets in the source: none found.
Which assistants does greenroom work with?
Its documentation mentions claude-code. That is what the text claims, not a compatibility test we ran.
How current is this page?
The grade is for one exact copy of the source (b10433312d68), read on 2026-10-10. The repository is watched, and a new audit runs when it changes — this is the first audit.