CLAWSTIN MORNING PAPER — 2026-03-13

Friday, 2026-03-13

INNOVATIONS

RESEARCHER

Researcher Report — 2026-03-13

Run time: 2026-03-13 01:01 ET Agent: Researcher (Haiku scan / Sonnet eval) Budget: Budget: $0.0036 / $5.00 used (15 calls, 2711in + 362out tokens) | $4.9964 remaining


Phase 1: Tech Research


Sources scanned: 656 items across HN + RSS feeds Candidates after scoring: 15 CBL evaluated: 15

EAT (queued to fridge)

- [EAT] LLM-driven Multimodal Recommendation — _✅ queued_ - [EAT] Taming OpenClaw: Security Analysis and Mitigation of Autonomous LLM Agent Threats — _✅ queued_ - [EAT] Measuring AI Agents' Progress on Multi-Step Cyber Attack Scenarios — _✅ queued_ - [EAT] Scaling Laws for Educational AI Agents — _✅ queued_ - [EAT] Governing Evolving Memory in LLM Agents: Risks, Mechanisms, and the Stability and Safety Governed Memory (SSGM) Framework — _✅ queued_

HOLD (notable but not fridged)

- [HOLD] WebWeaver: Breaking Topology Confidentiality in LLM Multi-Agent Systems with Stealthy Context-Based Inference — - [HOLD] Summarize Before You Speak with ARACH: A Training-Free Inference-Time Plug-In for Enhancing LLMs via Global Attention Reallocation — - [HOLD] Human-Centred LLM Privacy Audits: Findings and Frictions — - [HOLD] OrchMLLM: Orchestrate Multimodal Data with Batch Post-Balancing to Accelerate Multimodal Large Language Model Training — - [HOLD] Exponential-Family Membership Inference: From LiRA and RMIA to BaVarIA — - _(and 5 more HOLD items)_


Phase 2: PaperTrader Experiments


_No snapshot data available for today._

Phase 2 Errors

- ⚠️ No snapshot for today — cannot analyze performance


Phase 3: Optimization Analysis


> _Stale files and cron health are auditor territory (autoaudit). This phase covers cost and model routing only._

Session Model Usage (49 sessions, last 7d)


| Model | Mentions | Share | |-------|----------|-------| | Opus | 78 | 50% | | Sonnet | 50 | 32% | | Haiku | 28 | 18% |
Opus-heavy sessions: - `2026-03-12-session-26.md` (4x Opus) — Session 26 — 2026-03-12 (00:00–01:46 EDT, Opus) - `2026-03-11-session-25.md` (3x Opus) — Session 25 — 2026-03-11 (22:03–23:57 EDT, Opus) - `2026-03-11-session-20.md` (4x Opus) — Session 20 — 2026-03-11 (11:17–13:27 EDT, Opus) - `2026-03-11-session-18.md` (4x Opus) — Session 18 — 2026-03-11 (06:26–09:11 EDT, Sonnet→Opus) - `2026-03-10-session-14.md` (3x Opus) — Session 14 — 2026-03-10 (16:56–18:29 EDT, Sonnet→Opus)

Researcher Budget History


- last run: $0.0036 / $5.00 (0% utilized, 15 API calls)

Cost Optimization Opportunities


- Opus referenced in 78 mentions across 49 sessions (50% of model refs) → Review Opus-heavy sessions — most tasks could run on Sonnet at ~10x lower cost _Up to ~10x on affected calls_ - Researcher used only $0.0036 of $5.00 cap (0% utilization) → Consider reducing budget_cap_usd or adding more Phase 1/2 analysis depth _N/A — currently under-utilized_


Budget Summary


Total spent: $0.0036 / $5.00 cap API calls: 15 Tokens: 2711 input + 362 output
| Model | Input | Output | Cost | Note | |-------|-------|--------|------|------| | claude-haiku-4-5 | 185 | 25 | $0.000248 | CBL:WebWeaver: Breaking Topology Confide | | claude-haiku-4-5 | 195 | 25 | $0.000256 | CBL:Summarize Before You Speak with ARAC | | claude-haiku-4-5 | 175 | 23 | $0.000232 | CBL:Human-Centred LLM Privacy Audits: Fi | | claude-haiku-4-5 | 190 | 28 | $0.000264 | CBL:OrchMLLM: Orchestrate Multimodal Dat | | claude-haiku-4-5 | 167 | 27 | $0.000242 | CBL:LLM-driven Multimodal Recommendation | | claude-haiku-4-5 | 180 | 23 | $0.000236 | CBL:Exponential-Family Membership Infere | | claude-haiku-4-5 | 178 | 25 | $0.000242 | CBL:Taming OpenClaw: Security Analysis a | | claude-haiku-4-5 | 186 | 21 | $0.000233 | CBL:Security-by-Design for LLM-Based Cod | | claude-haiku-4-5 | 189 | 24 | $0.000247 | CBL:Preparing for the Quantum Era: Post- | | claude-haiku-4-5 | 172 | 23 | $0.000230 | CBL:Measuring AI Agents' Progress on Mul | | claude-haiku-4-5 | 179 | 26 | $0.000247 | CBL:The Unlearning Mirage: A Dynamic Fra | | claude-haiku-4-5 | 181 | 23 | $0.000237 | CBL:RewardHackingAgents: Benchmarking Ev | | claude-haiku-4-5 | 164 | 22 | $0.000219 | CBL:Scaling Laws for Educational AI Agen | | claude-haiku-4-5 | 191 | 24 | $0.000249 | CBL:Governing Evolving Memory in LLM Age | | claude-haiku-4-5 | 179 | 23 | $0.000235 | CBL:Social, Legal, Ethical, Empathetic a |

AUTO AUDIT RESULTS

AUTOAUDIT Summary -- 2026-03-13


Findings


CRITICAL


1. morning-brief cron timed out. Job `morning-brief` (7858a633) last run at 2026-03-12T22:33Z errored: "job execution timed out" after 420s. consecutiveErrors: 1. Ghost's 6 AM brief will not deliver tomorrow unless the underlying issue is resolved. Likely cause: Sonnet taking too long reading multiple state files + paper trader JSON. Consider increasing timeout or reducing scope.

WARNING


1. vital-widget-freshness smoke test FAIL. `vital-widget.json` file is missing entirely. Smoke test reports "File missing." The vital-widget-sync LaunchAgent was removed by Ghost in session 27 and stale files trashed — this is expected. However, smoke_test.py still checks for the file. The smoke test itself needs updating to remove this check, since the widget sync is intentionally decommissioned.
2. 2 cron jobs in error state: morning-brief, clear-chronic-wednesday. morning-brief covered in CRITICAL #1. `clear-chronic-wednesday` (1fe32258) error: "Delivering to Signal requires target <E.164|uuid:ID|group:ID|signal:group:ID|signal:+E.164>" — the delivery target `+15406208059` is set but the job's own message tries to send a Signal DM directly instead of relying on announce delivery. The `model` was fixed to haiku in session 27, but the delivery routing error persists. consecutiveErrors: 1.
3. balance-anchor smoke test WARN. Reports "no last_checked timestamp" but `vital-balance-anchor.json` contains `"last_checked": "2026-03-13T01:00:00Z"`. Balance: $115.02. This is a smoke_test.py bug — carried forward from last audit. The anchor itself is healthy.
4. vitals-api and vital-server LaunchAgents exit -15 (SIGTERM). `clawstin.vital-server` (pid 75085, exit -15) and `clawstin.vitals-api` (pid 95401, exit -15). Smoke test confirms flask-api on :8765 IS responding. Both show SIGTERM — launchd restarts. No functional impact, but 2nd consecutive audit with these signals.
5. com.clawstin.balance-notify LaunchAgent exit code 1. Pre-audit flagged this agent with exit status 1 (failure). Not a new finding but worth monitoring.
6. Stale FER plist in lifeboat. `lifeboat-system/launch-agents/clawstin.fer-monitor.plist` still exists, pointing to deleted `scripts/fer-monitor.py`. dependency_map.json and balance-scrape-cron.md were cleaned in session 27, but the lifeboat plist copy was missed. The plist is not loaded (not in launchctl list), so no functional impact — just a dead artifact.
7. Context load: 2,163 words (threshold: 1,500). Over threshold. Files over 400-word threshold: - AGENTS.md: 772 words (auto-injected) - memory/2026-03-12.md: 542 words (startup-read)
Full breakdown: AGENTS.md 772, SOUL.md 64, TOOLS.md 171, IDENTITY.md 32, USER.md 32, HEARTBEAT.md 95, MEMORY.md 117, STYLE.md 17, SYNC.md 103, WORKING_MEMORY.md 22, memory/2026-03-12.md 542, memory/2026-03-11.md 196. Total: 2,163.
Improvement from last audit (2,769 → 2,163, -606 words) due to memory/2026-03-11.md distillation (1,297 → 196). AGENTS.md remains the largest single file at 772 words — consider whether any sections could move to on-demand reads.
8. SCHEDULE.md has 3 past-due entries from 2026-03-11. These were noted as "resolved/removed" in session 27 daily log but are still present in SCHEDULE.md. Either the cleanup didn't persist or the file wasn't saved. Entries: Tori post office (3/11 10:00), Garry Dan Call (3/11 13:00), Hound BAM (3/11 15:54).

Carried Over


1. SCHEDULE.md sort order + stale entries — 8th consecutive audit. The 3/11 entries were logged as removed in session 27 but remain. 2. MEMORY.md pending items from 2026-03-04 — Ghost closed these in session 27 ("doesn't belong in MEMORY.md"). Resolved — dropping from carry-over. 3. Stale FER references — Mostly resolved in session 27. Only `lifeboat-system/launch-agents/clawstin.fer-monitor.plist` remains. 5th consecutive (partial). 4. smoke_test.py false positives — balance-anchor "no last_checked" and vital-widget-freshness (checking for intentionally removed file). 2nd consecutive.

Past-Due Schedule Entries


- `2026-03-11 10:00 -- Tori post office help` - `2026-03-11 13:00 -- Garry Dan Call` - `2026-03-11 15:54 -- Hound BAM`

Fired One-Shot Reminders


- No fired one-shot plists detected. `clawstin.ats.dentist-may-28th-8am` loaded but not yet fired (fires May 28). - `balance-burn-calibration-review` (one-shot, deleteAfterRun) scheduled for 2026-03-15T09:00Z — not yet fired.

Step Completion Checklist

Step 1 -- Pre-Audit Data: completed (11 checks; 0 errors, 2 warnings: LaunchAgents flagged, schedule past-due) Step 1.5 -- Smoke Tests: completed (12 checks; 9 pass, 2 warn, 1 fail: vital-widget missing, balance-anchor field mismatch, 2 cron errors) Step 2 -- Last Report Review: completed (reviewed 2026-03-12 report; morning-brief resolution reversed — now in error again; MEMORY.md pending items resolved by Ghost) Step 3 -- Daily Integration: completed (2026-03-12 log reviewed; no 2026-03-13 log yet; all referenced scripts verified at stated paths) Step 4 -- Git Diff + Downstream: completed (5 commits reviewed; watchdog.py added and indexed; vital-widget files cleaned; no stale references to changed values found in configs) Step 5 -- File Health Review: completed (MEMORY.md 117w healthy; context load 2,163w over 1,500 threshold; AGENTS.md 772w over 400 threshold; SCHEDULE.md has 3 stale past-due entries; all injected files checked — no contradictions or inaccuracies found) Step 6 -- Cron + Automation: completed (22 cron jobs reviewed; morning-brief timeout error, clear-chronic-wednesday delivery error; model assignments appropriate: Opus for autoaudit, Sonnet for morning-brief/security-guard, Haiku for all routine jobs; bobby-litter-check correctly uses systemEvent/main) Step 7 -- Script Validation: completed (send-todo.sh ✓, triage-proton.py ✓, triage.py ✓, watchdog/ directory ✓ with 6 files; all cron-referenced scripts verified: hue.py, fridge-discovery.py, zero_token_heartbeat.py, morning-paper.py, openclaw-update.py, rent-reminder.py, researcher.py, all 5 trading signal monitors ✓) Step 8 -- Cross-File Consistency: completed (stale FER plist in lifeboat; SCHEDULE.md entries logged as removed but still present; trial.py still references fer-monitor.py as an example in help text — cosmetic only; scripts/balance-scrape-cron.md is clean reference doc, no issues)

CAPABILITY QUEUE

Queue is empty.

PAPER TRADING

Model Portfolio Value P/L Cash Holdings
M01 Momentum Chaser $991.00 $-9.00 (-0.9%) $241.02 MSTR 1.1148sh @$137.37, ADBE 0.5523sh @$270.13, QQQ 0.2491sh @$597.34, MSFT 0.3659sh @$401.89, CRM 0.7619sh @$199.24
M02 Trend Follower $998.40 $-1.60 (-0.2%) $850.00 NVDA 0.8109sh @$183.01
M03 Momentum Rotator $1159.43 +$159.43 (+15.9%) $1159.43 Cash only
M04 Headline Trader $1000.00 +$0.00 (+0.0%) $1000.00 Cash only
M05 Hot Sector Rotator $1000.00 +$0.00 (+0.0%) $1000.00 Cash only
M06 Earnings Anticipator $1000.00 +$0.00 (+0.0%) $1000.00 Cash only
M07 Panic Buyer $1000.00 +$0.00 (+0.0%) $1000.00 Cash only
M08 Smart Money Tracker $1000.00 +$0.00 (+0.0%) $1000.00 Cash only
M09 Gap Fader $1000.00 +$0.00 (+0.0%) $1000.00 Cash only
M10 Coil & Breakout $1000.00 +$0.00 (+0.0%) $1000.00 Cash only

Live prices: AAPL: $255.75, ADBE: $270.13, CRM: $199.24, GOOGL: $303.51, MSFT: $401.89, MSTR: $137.37, NVDA: $183.01, QQQ: $597.34, SPY: $666.04, TSLA: $395.01, WDAY: $134.15

SECURITY AUDIT

Security Guard Report - 2026-03-12

Patrol time: 03:30 AM (America/New_York) Agent: Security Guard (Claude Opus) AutoAudit verified: Yes (2026-03-12) Previous report: 2026-03-11


Executive Summary


Overall threat level: MEDIUM (unchanged from yesterday)
No new critical findings. All prior findings persist with minimal change. Key observations this patrol:
- lifeboat-system/ credential directory - 5th consecutive patrol. Still in workspace with full API key suite. Individual files are 600, but several parent subdirectories (den/, openclaw-config/, rclone/, signal-data/) are 755 (world-readable directory listing). No change in risk profile. - vitals-api down - Both clawstin.vital-server and clawstin.vitals-api LaunchAgents show exit -15 (SIGTERM). No process is listening on port 8765. The api.clawstin.org Cloudflare Tunnel route returns connection refused. This is a reliability issue, not a security issue. - No new credential exposure - Hellbot subagent files (new since last patrol) contain no credentials. No .env files found. No stray API keys detected. - LuLu firewall running - Both app (PID 1491) and system extension (PID 699) active. - No external threat intelligence - 5th consecutive patrol unable to check NVD/Node.js/macOS advisories (sandbox has no network access).
This patrol identified 8 findings: 1 severity 7, 3 severity 5, 1 severity 3, and 3 severity 1-2.


FINDINGS


SG-2026-03-12-001: Lifeboat-System Credential Directory in Workspace

Severity: 7 (unchanged - 5th consecutive patrol)
Evidence: ``` $ ls -la /Users/aicomputer/.openclaw/workspace/lifeboat-system/ drwxr-xr-x 9 aicomputer staff 288 Mar 6 17:21 . (755 parent dir) drwx------ 5 aicomputer staff 160 Mar 6 17:21 cloudflared (700 OK) drwxr-xr-x 4 aicomputer staff 128 Mar 6 17:21 den (755 BAD) drwx------ 7 aicomputer staff 224 Mar 6 17:21 gmail-tokens (700 OK) drwxr-xr-x 67 aicomputer staff 2144 Mar 11 09:15 launch-agents (755, no creds) drwxr-xr-x 3 aicomputer staff 96 Mar 6 17:21 openclaw-config (755 BAD) drwxr-xr-x 3 aicomputer staff 96 Mar 6 17:21 rclone (755 BAD) drwxr-xr-x 5 aicomputer staff 160 Mar 6 17:21 signal-data (755 BAD) ```
Credential files inside are all -rw------- (600), but 4 of 7 subdirectories are 755 - any local user can traverse and ls the directory contents (seeing filenames), though they cannot cat the file contents.
Contents confirmed present (file permissions are 600): - den/fernet-key.b64 - 61 bytes, Fernet master encryption key - den/creds.enc - 4792 bytes, encrypted credential store (key is adjacent) - openclaw-config/openclaw.json - plaintext API keys: sk-ant-[REDACTED], sk_[REDACTED] (ElevenLabs), sk-or-[REDACTED] (OpenRouter) - gmail-tokens/credentials.json - Google OAuth client_secret: GOCSPX-[REDACTED] - gmail-tokens/token-*.json - 3 Gmail OAuth tokens with access + refresh tokens - rclone/rclone.conf - Google Drive OAuth tokens - signal-data/535318 - Signal key material - cloudflared/2c29ad40-*.json - Cloudflare tunnel secret
Mitigating factors: - All individual files are 600 (owner-read-only) - lifeboat-system/ is in .gitignore - not committed - No git remote configured (local-only repo, confirmed via .git/config) - Nightly backup ran successfully at 02:06 today (encrypted zip to Google Drive)
Remaining risk: Fernet key stored as a file next to the encrypted store negates encryption benefit. Directory permissions on den/, openclaw-config/, rclone/, signal-data/ should be 700.
Rubric: Severity 7 - credential storage locally accessible, not externally exposed.
Action: chmod 700 on den/, openclaw-config/, rclone/, signal-data/. Consider moving lifeboat-system/ out of workspace. Store Fernet key in macOS Keychain only.


SG-2026-03-12-002: Cloudflare Tunnel - 4 Routes, 2 Active, 1 Unauthenticated

Severity: 5 (unchanged - matches accepted risk SG-2026-03-10-002 for static site)
Evidence: ``` $ cat lifeboat-system/cloudflared/config.yml ingress: - hostname: clawstin.org -> localhost:8877 (Python http.server, static, NO AUTH) ACTIVE - hostname: voice.clawstin.org -> localhost:3334 (NOT RUNNING) - hostname: webhook.clawstin.org -> localhost:18789 (NOT RUNNING) - hostname: api.clawstin.org -> localhost:8765 (NOT RUNNING - vitals-api SIGTERM) - service: http_status:404 (catch-all)
$ ps aux | grep "8877" PID 75085: Python -m http.server 8877 --directory /Users/aicomputer/clawstin-site
$ ps aux | grep "8765" (no process found)
$ launchctl list | grep vital 75085 -15 clawstin.vital-server (SIGTERM exit) 62742 -15 clawstin.vitals-api (SIGTERM exit) ```
Analysis: - clawstin.org (8877): Static site, no auth - accepted risk (SG-2026-03-10-002) - voice.clawstin.org (3334): Not running. No risk. - webhook.clawstin.org (18789): Not running. No risk. - api.clawstin.org (8765): Down (SIGTERM). No security risk while down.
Rubric: Severity 5 - static site serves display telemetry only (accepted). API is down.


SG-2026-03-12-003: Chrome Remote Desktop Running

Severity: 3 (accepted risk - SG-2026-03-09-005, permanent accept)
Evidence: ``` $ ps aux | grep remoting PID 1505: remoting_me2me_host --host-config=...org.chromium.chromoting.json --ssh-auth-sockname=/tmp/chromoting.aicomputer.ssh_auth_sock PID 1478: remoting_me2me_host_service --run-from-launchd PID 805: remoting_agent_process_broker (root) ```
Running since Saturday boot. Full desktop access via Google account + PIN.
Rubric: Accepted risk (permanent). Downrated to 3 per deduplication rule.


SG-2026-03-12-004: Signal-CLI HTTP Daemon on Localhost (No Auth)

Severity: 5 (unchanged)
Evidence: ``` $ ps aux | grep signal PID 88391: java ... org.asamk.signal.Main -a +16072208785 daemon --http 127.0.0.1:8080 --no-receive-stdout ```
signal-cli daemon on 127.0.0.1:8080 - HTTP, no TLS, no authentication. Any local process can send Signal messages as the registered number.
Mitigating factors: - Bound to 127.0.0.1 only (not 0.0.0.0) - Requires local process access - LuLu firewall active
Rubric: Severity 5 - localhost service without auth, requires local access.


SG-2026-03-12-005: Proton Bridge Sentry DSN in Process Table

Severity: 2 (unchanged - informational)
Evidence: ``` $ ps aux | grep proton PID 1508: crashpad_handler ... sentry_key=ea31df[REDACTED] ```
Sentry DSN is a public ingest-only key (by design). Not a private API credential.
Rubric: Severity 2 - informational.


SG-2026-03-12-006: Test Password in Fixture File

Severity: 1 (unchanged - informational)
Evidence: ``` $ grep "password" scripts/test-llc-intake.json "password": "SecureP@ss1" ```
Test fixture with dummy password. File is 600. Not a real credential.
Rubric: Severity 1 - test data.


SG-2026-03-12-007: Brave Browser Remote Debugging Port Active

Severity: 2 (informational - unchanged)
Evidence: ``` Multiple Brave Browser Helper processes with --remote-debugging-port=18800 --user-data-dir=/Users/aicomputer/.openclaw/browser/openclaw/user-data ```
OpenClaw headless browser for automation. Localhost-only, separate profile.
Rubric: Severity 2 - localhost dev tool, required for operation.


SG-2026-03-12-008: Ollama Local LLM Server Running

Severity: 1 (informational - new finding)
Evidence: ``` $ ps aux | grep ollama PID 1511: /Applications/Ollama.app/Contents/Resources/ollama serve Running since Sat 07AM ```
Ollama local LLM inference server. Default bind: 127.0.0.1:11434.
Rubric: Severity 1 - localhost-only local LLM. No credential risk.


Threat Landscape - External Intelligence


Status: UNAVAILABLE (5th consecutive patrol)
Sandbox has no network access. Unable to check NVD, Node.js advisories, macOS bulletins, OpenClaw GitHub, or Signal CLI issues.
Recommendation: Ghost should periodically check: 1. https://nvd.nist.gov for CVEs affecting macOS Sequoia, Python 3.9, Java 21 2. https://support.apple.com/en-us/100100 for macOS security updates 3. https://github.com/AsamK/signal-cli/releases for signal-cli 0.13.24 vulnerabilities


Risk Acceptance Cross-Reference


| Finding | Matches Accepted Risk? | Action | |---------|----------------------|--------| | SG-001 (lifeboat creds) | No - not yet accepted | Include in brief | | SG-002 (Cloudflare tunnel) | Yes - SG-2026-03-10-002 | Noted, not escalated | | SG-003 (Chrome Remote Desktop) | Yes - SG-2026-03-09-005 | Downrated to 3 | | SG-004 (signal-cli no auth) | No | Include in brief | | SG-005 (Proton sentry DSN) | No (low sev) | Informational | | SG-006 (test password) | No (low sev) | Informational | | SG-007 (Brave debug port) | No (low sev) | Informational | | SG-008 (Ollama) | No (low sev) | Informational |


Changes Since Last Patrol (2026-03-11)


1. Vitals-API down - Port 8765 has no process. LaunchAgents exited SIGTERM. Availability issue, not security. 2. Hellbot subagent deployed - New hellbot/ directory. No credentials found. state.enc is encrypted. cli.py is 600. 3. New commands added - FIX.md, BOOK.md, BRE.md, SSA.md, END.md, ULP.md. No security implications. 4. Nightly backup successful - Committed 02:00, lifeboat uploaded 02:06. 5. Directory permission inconsistency persists - den/, openclaw-config/, rclone/, signal-data/ still 755. 6. No world-writable files - find confirmed zero results. 7. No .env files - find confirmed zero results.


File Permission Audit Summary


| Path | Permission | Expected | Status | |------|-----------|----------|--------| | ~/.openclaw/ | drwx------ | 700 | OK | | ~/.openclaw/credentials/ | drwx------ | 700 | OK | | ~/.openclaw/creds.enc | -rw------- | 600 | OK | | ~/.openclaw/openclaw.json | -rw------- | 600 | OK | | workspace/.gitignore | -rw------- | 600 | OK | | lifeboat-system/ | drwxr-xr-x | 700 | BAD (755) | | lifeboat-system/den/ | drwxr-xr-x | 700 | BAD (755) | | lifeboat-system/cloudflared/ | drwx------ | 700 | OK | | lifeboat-system/gmail-tokens/ | drwx------ | 700 | OK | | lifeboat-system/openclaw-config/ | drwxr-xr-x | 700 | BAD (755) | | lifeboat-system/rclone/ | drwxr-xr-x | 700 | BAD (755) | | lifeboat-system/signal-data/ | drwxr-xr-x | 700 | BAD (755) | | All credential FILES inside | -rw------- | 600 | OK |


Summary Table


| ID | Finding | Severity | Trend | Accepted? | |----|---------|----------|-------|-----------| | SG-001 | Lifeboat-system creds in workspace | 7 | unchanged (5th) | No | | SG-002 | Cloudflare tunnel (4 routes) | 5 | unchanged | Partial | | SG-003 | Chrome Remote Desktop | 3 | unchanged | Yes | | SG-004 | signal-cli HTTP no auth | 5 | unchanged | No | | SG-005 | Proton sentry DSN in ps | 2 | unchanged | No | | SG-006 | Test password in fixture | 1 | unchanged | No | | SG-007 | Brave remote debug port | 2 | unchanged | No | | SG-008 | Ollama local LLM server | 1 | NEW | No |
No findings rated 9-10. No Signal alert required.