eed04d55a5
T9 (burn-before-enqueue, #908 arrow #1) forced the durable pending-write failure via `chmod 500 "$STATE_DIR"`. That works only as a non-root user: DAC permission checks (including a read-only directory) are bypassed entirely by root, and Woodpecker CI containers run as root — so the write silently succeeded, observed_seq advanced, and all three "failed-write" assertions went spuriously green. The chmod also blocked the cursor's (observed_seq) own mktemp in the same directory, so even under a non-root run T9 could never have caught a premature cursor commit — narrower coverage than its comment claimed. The product code (store.sh) was already correct: the durable pending write is textually first and the cursor write is gated on its success (verified independently, unchanged here). Fix: bind-mount a throwaway file over $STATE_DIR/pending.jsonl (the exact _atomic_write rename TARGET for the durable write) so the atomic write's `mv -f tmp target` hits EBUSY — the kernel refuses to rename any file onto an ACTIVE MOUNT POINT. That is a structural VFS constraint, not a DAC check: no uid, root included, can bypass it short of an explicit umount. observed_seq is left an ordinary writable file, untouched by the mount, so a regression that commits the cursor before/independent of the durable write is now genuinely caught. The mount + forced enqueue run inside a private `unshare --mount --user --map-root-user` namespace so the same mechanism applies whether the harness runs unprivileged (dev) or as root (CI); the bind mount is private to that namespace and vanishes the instant it exits, so no manual cleanup step is needed afterward. sweep: grep chmod/chattr across all packages/mosaic/framework/tools/wake/test-*.sh -> T9's `chmod 500 "$STATE_DIR"` (fixed here) was the ONLY permission-based fault-injection use. Every other chmod hit is `chmod +x` making a mock/ fixture script executable (benign setup, not failure injection, unaffected by root): test-wake-reconcile.sh:68, test-wake-detector.sh:110, test-wake-digest-hmac.sh:164, test-wake-beacon.sh:70/79/88/103/311, test-wake-store-ack.sh:246 (mock curl/wget/nc/ssh binaries). Verified: enqueue under the injection exits non-zero as the current non-root user, with the process reporting uid 0 INSIDE the namespace (genuine kernel-mapped root) and still failing EBUSY — a direct proof the mechanism is root-proof, not just an argument. fakeroot was tried per request but only fakes stat/chown (no real mount capability), so it is not a meaningful vehicle for this injection and was not used as the proof. Coverage-RED reproduction: temporarily moved the observed_seq cursor write in store.sh to before the durable pending write (premature commit) -> T9 went RED (observed_seq wrongly advanced to 2, post-failure allocation wrongly resumed at 3); reverted -> GREEN again, diff-clean against origin. Full `pnpm run test:framework-shell` rc=0 (all harnesses, T9 included, 11/11 wake store/ack groups green); re-run with openssl masked from PATH (command -v openssl confirmed empty) rc=0, wake store/ack harness unaffected (an unrelated harness, test-wake-digest-hmac.sh, gracefully SKIPs without openssl -- pre-existing, orthogonal, HMAC is W3/A5 not W2). shellcheck 0.11 clean (one deliberate SC2016 on nested-shell variable expansion, disabled inline per this repo's existing convention). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0158NZqN2n2ymKFeJAZ4GUCb
@mosaicstack/mosaic
CLI package for the Mosaic self-hosted AI agent platform.
Usage
mosaic wizard # First-run setup wizard
mosaic gateway install # Install the gateway daemon
mosaic config show # View current configuration
mosaic config hooks list # Manage Claude hooks
Headless / CI Installation
Set MOSAIC_ASSUME_YES=1 (or ensure stdin is not a TTY) to skip all interactive prompts. The following environment variables control the install:
Gateway configuration (mosaic gateway install)
| Variable | Default | Required |
|---|---|---|
MOSAIC_STORAGE_TIER |
local |
No |
MOSAIC_GATEWAY_PORT |
14242 |
No |
MOSAIC_DATABASE_URL |
(none) | Yes if tier=team |
MOSAIC_VALKEY_URL |
(none) | Yes if tier=team |
MOSAIC_ANTHROPIC_API_KEY |
(none) | No |
MOSAIC_CORS_ORIGIN |
http://localhost:3000 |
No |
Admin user bootstrap
| Variable | Default | Required |
|---|---|---|
MOSAIC_ADMIN_NAME |
(none) | Yes (headless) |
MOSAIC_ADMIN_EMAIL |
(none) | Yes (headless) |
MOSAIC_ADMIN_PASSWORD |
(none) | Yes (headless) |
MOSAIC_ADMIN_PASSWORD must be at least 8 characters. In headless mode a missing or too-short password causes a non-zero exit.
Example: Docker / CI install
export MOSAIC_ASSUME_YES=1
export MOSAIC_ADMIN_NAME="Admin"
export MOSAIC_ADMIN_EMAIL="admin@example.com"
export MOSAIC_ADMIN_PASSWORD="securepass123"
mosaic gateway install
Runtime launchers
mosaic claude # Launch Claude Code with Mosaic injection
mosaic yolo claude # …with --dangerously-skip-permissions
mosaic codex | opencode | pi
mosaic claudex (EXPERIMENTAL)
Runs GPT models inside the Claude Code harness by pointing Claude Code at a
local claude-code-proxy that
translates the Anthropic Messages API to a ChatGPT-subscription (Codex OAuth)
backend. This is not Anthropic Claude — model behavior, tool use, and output
quality may differ. Intended for evaluation, not production delivery.
mosaic claudex # launch (prompts through the proxy readiness gate)
mosaic yolo claudex # …with --dangerously-skip-permissions
mosaic claudex --print "hello" # trailing args are forwarded to Claude Code
Prerequisite: the claude-code-proxy binary must be installed and
authenticated (claude-code-proxy codex auth …). mosaic claudex runs a
preflight that verifies the binary, the OAuth state (triggering a device re-auth
if needed), and a trusted local listener before launching; it fails closed
if the proxy cannot be brought up with a verified identity.
Isolation (never touches your real Claude state). claudex always launches
against an isolated CLAUDE_CONFIG_DIR (default ~/.config/mosaic/claudex/home).
The ambient CLAUDE_CONFIG_DIR is deliberately ignored, and a guard proves the
resolved dir can never be — or live under — the real ~/.claude. A claudex
session therefore cannot mutate your normal Claude Code config.
No token leakage. claudex never reads the proxy's credential file. Claude
Code is handed only ANTHROPIC_AUTH_TOKEN=unused pointed at the loopback proxy;
the entire credential-bearing env family (ANTHROPIC_*, AWS_*, GOOGLE_CLOUD_*,
GOOGLE_APPLICATION_CREDENTIALS, *_TOKEN, *_KEY, *_SECRET, …) is stripped
from the composed environment. The Bedrock/Vertex routing switches
(CLAUDE_CODE_USE_BEDROCK, CLAUDE_CODE_USE_VERTEX, and the _SKIP_*_AUTH
pair) are force-removed regardless of value — otherwise their mere presence
would route Claude Code to the real Anthropic API via AWS/GCP and bypass the
proxy. The proxy holds the real OAuth credential.
Model tiers (override via env).
| Tier | Env var | Default |
|---|---|---|
| primary (opus/sonnet) | ANTHROPIC_MODEL |
gpt-5.6-sol |
| small/fast (haiku) | ANTHROPIC_SMALL_FAST_MODEL |
gpt-5.6-luna |
Operator-provided values win over the defaults. Additional overrides:
MOSAIC_CLAUDEX_CONFIG_DIR (isolated config dir), ANTHROPIC_BASE_URL (proxy
endpoint).
Hooks management
After running mosaic wizard, Claude hooks are installed in ~/.claude/hooks-config.json.
mosaic config hooks list # Show all hooks and enabled/disabled status
mosaic config hooks disable PostToolUse # Disable a hook (reversible)
mosaic config hooks enable PostToolUse # Re-enable a disabled hook
Set CLAUDE_HOME to override the default ~/.claude directory.