Root causes found by reproducing a clean install from keigit:
1. PROFILE_PRIMS resolved only inside check_prereqs → unbound for
--no-execute (plan showed 0 prims for every profile) and silently
empty for --skip-prereqs. Now resolved unconditionally in install.sh
before any reader (SSoT).
2. Every profile (even minimal, advertised "no Rust compile") fell back
to a 5-15 min `cargo build --workspace` because no prebuilt release
binaries exist. Auto-set KEI_SKIP_RUST for profiles with no rust
primitives → minimal installs in ~18s (assembler only). cargo stays a
hard prereq because the agent assembler always compiles.
3. The assembler aborted the WHOLE install on any single bad manifest
(set -e). generate_agents is now tolerant: bad manifests print FAIL
but hooks/skills/settings still land. Commit-time validate stays strict.
4. Data bugs that broke the assembler:
- duplicate [taxonomy] table in _roles/{auditor,merger}.toml
- fal-ai-runner handoff → keimd-expert (not shipped in kit)
- infra-implementer-cicd forbidden_domain literal `${{ secrets.NAME }}`
collided with assembler ${{ }} placeholder detection
5. Metadata: KeiSei84 (nonexistent GitHub org) → KeiSeiLab/KeiSeiKit-1.0
across plugin manifests, bootstrap, README, docs, Cargo/npm metadata.
.claude-plugin/{plugin,marketplace}.json 0.16.0 → 0.38.0. SECURITY.md
supported version 0.14.x → 0.38.x.
feat: ship KeiSei tamagotchi statusline into the kit
- scripts/keisei-pet{,-update}.sh (portable, state under ~/.claude/pet/)
- install copies them to ~/.claude/scripts/
- settings-snippet adds statusLine (set-if-absent, never clobbers an
existing one) + 4 pet-update hooks (prompt/rust_write/github_block/sleep)
Verified: clean minimal install RC=0, zero FAIL, 38 agents + 52 hooks +
68 skills, settings valid, statusLine wired, pet renders, idempotent re-run.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
3.5 KiB
USB Exobrain — Linux Walkthrough
Platform-specific companion to
USB-BRAIN-GUIDE.md. Read the top-level guide first for prerequisites, warnings, and invariants.
On Linux, auto-mounted removable media typically lands at /media/$USER/<LABEL> (GNOME, KDE, auto-mounters) or /run/media/$USER/<LABEL> (systemd-udisks2). Substitute your actual mount point below.
1. Create the brain directory
BRAIN=/media/$USER/EXOBRAIN/my-brain
mkdir -p "$BRAIN"/{bin,memory,artifacts,manifests}
2. Download MCP server binaries
BASE=https://github.com/KeiSeiLab/KeiSeiKit-1.0/releases/download/v0.21.0
cd "$BRAIN/bin"
for n in darwin-arm64 darwin-x64 linux-x64 linux-arm64 windows-x64.exe; do
curl -fL -O "$BASE/kei-mcp-server-$n" 2>/dev/null || echo "skipped $n"
curl -fL -O "$BASE/kei-mcp-server-$n.sha256" 2>/dev/null || true
done
for f in kei-mcp-server-*.sha256; do sha256sum -c "$f"; done
chmod +x kei-mcp-server-linux-* kei-mcp-server-darwin-* 2>/dev/null || true
No Gatekeeper / xattr step on Linux. The chmod +x is still required — the executable bit is not restored by curl.
3. Write manifest.toml (schema v2)
cat > "$BRAIN/manifest.toml" <<'EOF'
[brain]
schema_version = 2
name = "my-brain"
created = "2026-04-22T00:00:00Z"
[paths]
memory = "memory/"
artifacts = "artifacts/"
manifests = "manifests/"
[paths.mcp_server]
darwin-arm64 = "bin/kei-mcp-server-darwin-arm64"
darwin-x64 = "bin/kei-mcp-server-darwin-x64"
linux-x64 = "bin/kei-mcp-server-linux-x64"
linux-arm64 = "bin/kei-mcp-server-linux-arm64"
windows-x64 = "bin/kei-mcp-server-windows-x64.exe"
EOF
4. Verify + attach
keisei list-adapters
keisei status # "no brain attached"
keisei attach "$BRAIN" --scope=user
Marker lands at ~/.keisei/attached.toml; Claude Code settings at ~/.claude/settings.json.
5. Verify in Claude Code
jq '.mcpServers["my-brain"]' ~/.claude/settings.json
6. Multi-client mount
keisei mount "$BRAIN"
Linux adapter paths: ~/.claude/settings.json, ~/.cursor/mcp.json, ~/.continue/config.json, ~/.config/zed/settings.json.
7. Project-scope
cd ~/path/to/your-repo
keisei attach "$BRAIN" --scope=project # claude-code + cursor only
8. Detach + unmount
keisei detach
umount /media/$USER/EXOBRAIN # or: sync && udisksctl unmount -b /dev/sdX1
If umount returns "target is busy", close any shell with its CWD under the mount, then retry. lsof +f -- /media/$USER/EXOBRAIN lists open handles.
Linux-specific troubleshooting
- Filesystem detection —
keiseicallsstatfs(2)at load time. exFAT (EXFAT_SUPER_MAGIC = 0x2011bab0) and FAT32 (MSDOS_SUPER_MAGIC = 0x4d44) trigger the SQLite-WAL-unsafe advisory. Format the USB as ext4 for reliable multi-client use. - Auto-mounted
noexecpartitions — some distros mount removable medianoexecby default. If the mcp server refuses to run, remount read-write-executable:sudo mount -o remount,exec /media/$USER/EXOBRAIN. Alternatively add a line to/etc/fstabkeyed by UUID (blkid /dev/sdX1). - Permissions drift — if you copy a brain from macOS via tar / rsync, the executable bit may not survive. Re-apply
chmod +x bin/kei-mcp-server-*. - Optional — systemd auto-attach — a
systemd-udevrule can runkeisei attach <mount>/my-brain --scope=userwhenever a labelled stick shows up, and a matchingudevremove rule can callkeisei detach. Out of scope for this guide; seeArchWiki: Udisks#Auto-mount.